The URL can be used to link to this page

Home My WebLink About20110527Duvall Di.pdfRECEIVED iUOMA Y 27 AM It: 06 BEFORE THE IDAHO PUBLIC UTILITIES COMMISSION IN THE MATTER OF THE ) APPLICATION OF ROCKY ) MOUNTAIN POWER FOR ) APPROVAL OF CHAGES TO ITS ) ELECTRIC SERVICE SCHEDULES ) AND A PRICE INCREASE OF $32.7 ) MILION, OR APPROXIMATELY )15.0 PERCENT ) CASE NO. PAC-E-l1-12 Direct Testimony of Gregory N. Duvall Redacted ROCKY MOUNTAIN POWER CASE NO. PAC-E-l1-12 May 2011 1 Q.Please state your name, business address and present position with Rocky 2 Mountain Power Company (the Company), a division ofPacifiCorp. 3 A.My name is Gregory N. Duvall. My business address is 825 NE Multnomah, Suite 4 600, Portland, Oregon, 97232. My present position is Director, Net Power Costs. 5 Qualifications 6 Q.Briefly describe your education and business experience. 7 A.I received a degree in Mathematics from University of Washington in 1976 and a 8 Masters of Business Administration from University of Portland in 1979. I was 9 first employed by PacifiCorp in 1976 and have held various positions in resource 10 and transmission planning, regulation, resource acquisitions and trading. From 11 1997 through 2000 I lived in Australia where I managed the Energy Trading 12 Departent for Powercor, a PacifiCorp subsidiar at that time. After retuing to 13 Portland, I was involved in direct access issues in Oregon and was responsible for 14 directing the analytical effort for the Multi-State Process ("MSP"). Curently, I 15 direct the work of the load forecasting group, the net power cost group, and the 16 renewable compliance area. 17 Purpose of Testimony 18 Q.What is the purpose of your testimony in this proceeding? 19 A.I present the Company's proposed net power costs ("NPC") based on actual data 20 for the 12-month period ending December 2010 with known and measurable 21 changes through December 2011 in support of the Company's 2011 test period in 22 this case. Specifically, my testimony: 23 . Addresses the specific adjustments related to the GRID model described in the REDACTED Duvall, Di- 1 Rocky Mountain Power 1 section for net power costs of the Commission Order No. 32196 ("2010 Rate 2 Case Order") in the Company's 2010 general rate case, Case No. PAC-E-1O- 3 07 ("2010 GRC"). 4 . Describes the major cost drvers in the 2011 NPC. 5 . Presents the Company's updated wind integration charges based on the 6 verifiable 2010 Wind Integration Study and explains how they are 7 incorporated in the curent fiing. 8 Summary of Net Power Costs in the Current Filng 9 Q. 10 A. 11 12 13 14 15 Q. 16 17 18 A. 19 20 21 22 23 What are the normalized net power costs for the test period? The normalized NPC for the 12 months ending December 2011 are approximately $82.8 millon on an Idaho allocated basis, or $1.312 bilion system-wide as presented in Exhibit No. 35. The allocation of total Company NPC to Idaho is presented in Exhibit No.2 in Company witness Mr. Steven R. McDougal's direct testimony. How do proposed NPC compare with the NPC that the Commission authorized in the Company's last general rate case, Case No. PAC-E-10-07? (the "2010 General Rate Case") The NPC authorized in the 2010 General Rate Case were $1.025 bilion on a total Company basis or $66.0 milion on an Idaho allocated basis. On a total Company basis, NPC have increased approximately $287 milion from $1.025 bilion to $1.312 billon. Idaho's allocated portion of NPC in the curent fiing is approximately $16.8 milion higher than the NPC curently included il customers' rates. REDACTED Duvall, Di- 2 Rocky Mountain Power 1 Q.Why have NPC increased? 2 A.To understand why NPC have increased, I present Table A which ilustrates the 3 changes in NPC by category: Table A Total Special Sales For Resale Total Purchased Power & Net Interchange Total Coal Fuel Bum Expense Total Gas Fuel Bum Expense Wheeling and Other ,Net system Load (4,518,537) 3,536,077 (3,434,792) (3,750,177) (151,448) 718,186rI~~~ ($347,22 ,266) ($1,826,473) ($12,281,065) $85,607,243 ($11,309,616) 4 As shown in Table A, the increase in NPC is drven largely by the unfavorable 5 change in revenues from sales for resale (wholesale sales), which is partially 6 offset by a favorable change in natual gas fuel bur expense. The decline in 7 wholesale sales revenues is primarily drven by a drop in market prices which are 8 30 percent lower in the curent filing as compared to the 2010 GRC fiing. In 9 addition, the Company had significantly less volume of wholesale sales and 10 thermal generation because: 1) low market prices make it more economic at times 11 to displace thermal units with purchases from the market; and 2) more capacity 12 from the thermal units is needed to provide reserves due to the 2,183 MW of wind 13 generation in the Company's Balancing Authority Areas ("BAA") at the end of 14 2010. 15 Q.Have you included the results of its most recent wind integration study into 16 the current filing? 17 A.Yes. In response to Order No. 32196 from the 2010 GRC, I am sponsorig a wid REDACTED Duvall, Di- 3 Rocky Mountain Power 1 2 3 4 5 6 7 8 9 10 11 12 13 Q. 14 15 A. 16 17 18 19 20 21 22 integration study ("Wind Study") that verifiably depicts the Company's wind integration costs. The inclusion of these costs is consistent with Order No. 32196, where the Commission stated: "We are not happy with this end result, because we believe these integration costs belong in base rates."l The Wind Study results showing a regulation reserve requirement of 533 megawatts ("MW") are fuher supported by the fact that the Company cared 629 MW of regulation reserves in 2010. Lastly, instead of the dollar per megawatt-hour ("MW") adder used in the last case, the Company improved its modeling and reflected the Wind Study's 533 MW regulation reserve requirement within the Generation and Regulation Initiative Decision ("GRID") model, resulting in an accurate depiction of how the Company's system incurs costs associated with wind generation resources in the test period. Please briefly explain the change in GRI associated with the Company's 2010 Wind Study. The results of the Wind Study showed that on average the Company would need to carr reserves for the intra-hour variations of load and wind of 533 MW. This is an increase of approximately 267 MW over the 2010 GRC fiing, which included regulation reserves to address load varability, but did not include any additional regulation reserves to manage the 2,183 MW of wind in the Company's BAAs, and instead addressed this reserve cost as an incremental. line item charge of $6.50/M. The $6.50/MWh charge for reserves in the 2010 GRC filing has been eliminated and replaced in the GRID model by the total wind net load ICase No. PAC-E-1O-07, Order No. 32196 page 30, Issued Februry 28, 2011. i"Regulation requirement" or "regulation reserves" consist of reserves available in 10 minutes (referred to as reguating reserves) and reserves available in 60 minutes (referred to as load following reserves). REDACTED Duvall, Di- 4 Rocky Mountain Power 1 2 3 4 5 6 7 8 Q. 9 10 A. 11 12 reserve requirement determined in the Wind Study. The GRID model then calculates the costs associated with the regulation reserves by backing off generation to supply those reserves that would otherwise be used to serve load or generate margins to reduce net power costs. These additional reserve requirements are one contrbutor to the reduction in generation seen in Table A of the Company's coal and natual gas-fired resources. I wil discuss this in more detail later in my testimony. How wil the normalized NPC approved by the Commission in this proceeding be used for setting retail rates in Idaho? The approved normalized NPC from this proceeding wil establish the level of NPC included in base rates and wil set the base NPC for purposes of the Energy Cost Adjustment Mechanism ("ECAM") and wil be tred-up to actual NPC 13 consistent with the mechanics of the ECAM. 14 Determination of NPC and Model Inputs and Outputs 15 Q. 16 A. Please explain NPC. NPC are defmed by the NPC report included as Exhibit No. 35 and include the 17 sum of fuel related expenses, wholesale purchase power expenses and wheeling 18 expenses, less wholesale sales revenue. 19 Q. 20 A. 21 Please explain how the Company calculates NPC. NPC are calculated using the GRID modeL. GRID is a production cost model that simulates the operation of the Company's power system on an hourly basis. REDACTED Duvall, Di- 5 Rocky Mountain Power 1 Q.Is the Company's general approach to the calculation of NPC using the 2 GRI model the same in this case as in previous cases? 3 A. 4 Q. 5 6 A. 7 Q. 8 A. Yes. The Company used the GRID model in its last rate filing in Idaho. Is the Company using the same version of the GRI model as used in the 2010 General Rate Case? Yes. What inputs were updated for this fiing? The system load, wholesale sales and purchase contracts for electrcity, natual 9 gas and wheeling, market prices for electrcity and natual gas, fuel expenses, 10 characteristics of the Company's generation facilities, normalized planned 11 outages and normalized forced outages of the Company's generation resources are 12 updated for this fiing. I address some of these components later in my testimony 13 along with changes to include transactions with the Cal ISO and changes in the 14 modeling of market caps. 15 Q. 16 A. What reports does the GRI model produce? The major output from the GRID model is the NPC report. This is attched to my 17 testimony as Exhibit No. 35. Additional data with more detailed analyses are also 18 available in hourly, daily, monthly and anual formats by heavy-load hours and 19 light-load hours. 20 Q. 21 A. Has the Company changed its modeling of normalized hydro generation? No. REDACTED Duvall, Di- 6 Rocky Mountain Power 1 Adjustments Identifed in the Commission Order 2 Q. 3 4 A. 5 6 7 8 9 Q. 10 11 A. 12 13 14 Q. 15 A. 16 17 18 19 20 21 22 Q. 23 24 A. What adjustments did the Commission direct the Company to make in Order No. 32196? In Order No. 32196, the Commission directed the Company to make adjustments for Normalization of Call Option Contracts, Wind Integration Costs, and Cal ISO Wheeling & Service Fees. The Normalization of Call Option Contracts the contracts that the Company has with Black Hils Power ("Black Hils") and the Sacramento Municipal Utilty Distrct ("SMUD"). How does the Company model the normalization of call option contracts in the current filing? For the contracts with Black Hils and SMUD, the Company modeled the energy delivery using a thee-year historic average, as directed by the Commission in Order No. 32196. Please discuss the Commission's decision on wind integration costs? In its order, the Commission denied the inclusion of wind integration costs in the Company's base net power costs, stating that: (n)o part to this case denies that wind integration costs are real costs; the consensus, however is that they cannot be readily forecast with accuracy, calculated or verified... The Company, we fmd, has not presented the Commission with a verifiable study depicting its wind integration costs.3 Has the Company updated its modeling approach in GRI to more accurately calculate wind integration costs? Yes. As part of its 2011 Integrated Resource Plan, the Company conducted an 3Case No. PAC-E-I0-07, Order No. 32196, page 30 issued Febru 28,2011. REDACTED Duvall, Di - 7 Rocky Mountain Power 1 2 3 4 5 Q. 6 7 A. 8 9 10 11 12 13 14 15 16 17 18 19 Q. 20 21 A. 22 23 extensive Wind Study that assessed the impact of integratig wind generation into its resource portfolio. The Wind Study is provided as Exhibit No. 36. In the curent fiing, the Company has reflected the results of the Wind Study, which I wil discuss in more detail later in my testimony. Has the Company included wind integration costs for wholesale transmission customers in its current NPC filing? Yes. The Company is required to integrate wholesale customers in a non- discriminatory manner under federal law pursuant to the Company's Open Access Transmission Tariff ("OATT"). The OATT does not allow the Company to charge for these integration costs separately from other charges under the OATT. However, customers are benefited by the Company being a balancing area authority and receiving wheeling revenues from wholesale customers which are applied as a credit against the cost of providing retail servce. On or before June 1, 2011, the Company wil file a wholesale rate case with the Federal Energy Regulatory Commission ("FERC") to update its wholesale transmission rates. As par of that filing, the Company is requesting a new Schedule 3A which, if approved by FERC, wil allow the Company to charge generators that do not have load in the BAA for regulation reserves. How does the Company respond to the Commission's decision on Cal iso fees in the 2010 Rate Case Order? The Commission's decision was made based on the fact that the Company's net power cost calculations in that case did not include transactions explicitly with the Cal iSO as the counterpart. The Commission indicated that transaction data REDACTED Duvall, Di- 8 Rocky Mountain Power 1 2 3 4 5 6 7 8 9 10 11 12 Q. 13 14 A. 15 16 17 18 19 20 21 22 should have been provided if the Company intended for the Cal iso fees to be a continuing expense. To address this concern, the Company added Cal iso transactions to the GRID modeL. The transactions with the Cal iso are a necessary part of the Company's activities to serve its load obligation and balance its system. Because of its diversity and liquidity, the Cal iso is an importnt counterpart for the Company when approaching time of delivery, which is also a fact that Monsanto agreed to in Mr. Mark T. Widmer's testimony: "Historical records reveal that most of the transactions with the Cal iso as a counter par are incured shortly before or on the actual day of delivery."4 As a result, the Company has modeled expected sales and purchase transactions with the Cal iso based on historical averages. Please explain how such transactions are modeled in the Company's current filing? Based on data in the same four-year historical period used to determne the market caps, the Company calculated the average amount of energy sold to and purchased from the Cal iso on a monthly basis and by heavy load hour ("HLH") and light load hour ("LLH"). The executed short term firm transactions that the Company included in the filing are through the end of March 2011. For the remainder of the test period, the Company modeled expected transactions with Cal iso at three major points of delivery based on historical information: Four Comers ("4C"), California Oregon Border ("COB") and Mona. Because these are expected transactions, similar to "System Balancing" sales and purchases, they 4See Docket No. ID PAC-E-I0-07, Confidential Corrected Direct Testimony of Mark Widmer, Page 24, Lines 1-2. REDACTED Duvall, Di- 9 Rocky Mountain Power 1 are grouped with the system balancing sales and purchases as modeled by GRID 2 at corresponding points of delivery. Together with these transactions, the 3 Company included the expected Cal iso wheeling fees and service fees. 4 Major Cost Drivers in the 2011 NPC 5 Q. 6 A. 7 8 9 10 11 12 13 14 15 16 17 Q. 18 A. 19 20 21 22 Please identify the major cost drivers in the 2011 NPC. As identified in my summary, the primary cost drver is the reduction in wholesale sales brought about by lower market prices and reduced thermal generation offset in par by a reduction in natual gas fuel expense. Other drvers include the following: . An increase in retail sales; · Expiration of a number of long-term wholesale power contracts; . Increased coal expense; and · An increase in the wholesale power and wind integration rates of the Bonnevile Power Administration ("BP A"). Finally, in this section I describe the treatment of the Monsanto contract and changes to modeling of market caps. Please describe the impact of lower wholesale sales volumes and revenues. Volumes of wholesale sales are down by 4.5 milion MWs as shown in Table 1. This reduced volume of wholesale sales is estimated to increases NPC by $157 millon. In addition, wholesale electrcity market prices are approximately 30 percent lower than those included in the 2010 GRC. Even with no change in volume, wholesale sales revenues would be approximately $190 milion lower REDACTED Duvall, Di - 10 Rocky Mountain Power 1 2 3 4 5 6 Q. 7 A. 8 9 10 Q. 11 12 A. 13 14 15 16 17 18 19 20 due to changes in market prices alone.5 The reduction in wholesale sales revenues is offset in part by lower natual gas fuel expenses in the curent fiing of $86 milion. Although purchased power prices are also lower, the overall purchased power expense increased by just under $2 milion due to increased volumes of 3.5 millon MWs. What are the major drivers to the reduced volume of wholesale sales? The primary drvers to reduced wholesale sales volume include reduced thermal generation, increased retail loads, and expiration of several long-term wholesale power contracts including purchase and sale exchanges. Why is thermal generation lower in the current filing as compared to the 2010GRC? As shown in Table A, coal fired generation is lower by 3.8 millon MW and natual gas fired generation is lower by 3.5 milion MW. The primar drvers for this reduction are economic displacement and increased reserves modeled in GRID. Economic displacement of thermal units is a consequence of market price reductions that result in a higher frequency of times when thermal generation is uneconomic to operate. Thermal generation is also lower due to the explicit modeling of regulation reserves in GRID, which requires thermal generation to be backed down to carr the necessary regulation reserves requied to integrate load and wind generation variability and maintain reliabilty. 519.2 milion megawatt-hours ("MW") at prices that are approximately $9.9/M lower. REDACTED Duvall, Di- 11 Rocky Mountain Power 1 Q. 2 3 A. 4 5 6 7 8 9 10 11 12 13 14 15 16 Q. 17 A. 18 19 20 21 22 Please explain how the drop in wholesale electricity market prices and natural gas prices affects how the GRI model holds reserves. The change in the wholesale electrcity market prices is greater than the drop in natul gas prices which reduces the amount of time the Company's natural gas- fired resources are economically dispatched. When the GRID model determines that it is uneconomic to dispatch a gas-fired resource and that resource was carring reserves, then that reserve carring requirement is shifted to other resources including coaL. For example, if Lake Side I is not dispatched because it is uneconomic to do so, it is no longer available to hold reserves, therefore the model wil hold reserves on the next available resource, which would show as a decline in generation of that resource due to the fact that it is now holding reserves rather than being able to generate and serve load or sell into the wholesale market. This decline in generation associated with reserve holding is apparent in the reductions in Table A of the Company's coal and natual gas-fied resources. Do the NPC in this filng reflect increases in coal costs since the 2010 GRC? Yes. As shown in Table A, NPC are higher by approximately $12 milion even though generation from coal units has declined by 3.4 millon MWs. Notwithstanding this decline in the volume of coal generation, increased costs of third-part coal supply and transportation agreements, and cost increases at the Company's captive mines are drving these costs. Details on coal costs are provided in the direct testimony of Company witness Ms. Cindy A. Crane. REDACTED Duvall, Di - 12 Rocky Mountain Power 1 Q. 2 A. 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 What are the major changes to power contracts in the test period? The contracts that have expired or wil expire, change or are new in the test period include: . On June 30, 2011, the exchange contract between the Company and the Alcoa Power Generating Inc. ("APGI") for approximately 1 00 MW of capacity from the Rocky Reach project expires. Under this contract, the Company receives energy durg peak periods and retus energy durng off-peak periods. . On October 31, 2011, the contract between the Company and the Chelan Public Utility Distrct ("Chelan PUD") for generation from the Rocky Reach project expires. Power purchased by the Company under this contract is priced at the embedded cost of the project. . On August 31, 2011, the contract between the Company and the Bonnevile Power Administration ("BPA") for 575 MW of capacity expires. Under this contract, the Company receives energy durg peak periods and retus energy during off-peak periods. In addition, power received under this contract is delivered directly to a variety of the Company's load pockets in the western area at the Company's discretion. . On September 30, 2011, the contract between the Company and the Grat Public Utilty Distrct ("Grant PUD") for displacement generation expires, which is priced at BPA's Priority Fir Power ("PF") rate. . On January 1, 2011, the amount of sales to the Public Service Company of Colorado ("PSCol") reduces per the contract terms, which is a legacy sales contract at relatively high contract prices. REDACTED Duvall, Di- 13 Rocky Mountain Power 1 Q. 2 A. 3 4 5 6 7 Q. 8 9 A. 10 11 12 13 14 15 16 17 Q. 18 19 A. 20 21 Does the fiing reflect an increase in load that impacts NPC? Yes. This fiing reflects an increase of approximately 1.2 percent in the total Company system load compared to loads reflected in the 2010 GRC. All else.held constant, increased load increases NPC, which in this case increased NPC by approximately $15 milion. The system load in this fiing is the actual load in 2010, adjusted for temperatue. What assumptions did the Company make in regard to the power rates and transmission rates proposed in the current rate cases of the BP A? The BP A rate cases wil determine the new rates for the fiscal period begining in October 2011. Given the curent proposals made by BP A, the Company assumes that the wheeling expenses of the Company's transmission contracts with BPA would not change in the new BP A rate effective period that begins in October 2011. In the curent filing, the Company has incorporated the proposed wind integration charge at $1.32/kilowatt(kW)-month begining in October 2011, which is a change from the curent $1.29/kW-month. The Company has also incorporated the impact of BP A's proposed charges for reserves and power. Does the Company expect to update the expenses related to all contracts with BPA? Yes. The Company wil update its NPC on rebuttal with the final decision of the BPA rate cases curently expected in July 2011, or when better information becomes available. REDACTED Duvall, Di- 14 Rocky Mountain Power 1 Q. 2 3 A. 4 5 6 7 8 9 Q. 10 A. 11 12 13 14 15 16 17 18 19 20 21 22 How is the price of the contract with Monsanto for interruptible products reflected in the test period? Given the interim agreement with Monsanto, the previous contract price at $IIW-month is effective through April 28, 2011. Beginning on April 29, 2011, the contract price is equivalent to $" milion per year as authorized by the Commission or $IIW-month for 162 MW for a term ending May 31, 2011. We expect that should the Company and Monsanto need additional time to complete a long term agreement, then this interim agreement wil be extended. Please describe the Company's change to the modeling of market caps? To address the issues around the Company's assumption about market caps durng the graveyard hours, the Company reviewed its overall approach to market caps and developed a more comprehensive approach to modeling market depth. Instead of specifying market depth for graveyard hours only, the Company now proposes to specify market depth durg all hours, segregated by HLH and LLH periods. The Company believes that a market may be liquid, but this liquidity does not translate into unlimited sales at any time of day or night. Due to load requirements and transmission constraints in the region and static assumptions about market prices in GRID, among other things, the Company may not be able to sell all its economic generation to the markets. The market depths for wholesale sales in GRID are now determined based on the historical short-term firm transactions during the same 48-month period on which availabilty of the thermal generation is based. The depths are then reduced by the quantity of short-term REDACTED Duvall, Di- 15 Rocky Mountain Power 1 firm transactions that the Company has included in the normalized NPC study for 2 the test period in all sales markets. 3 Wind Integration Costs 4 Q. 5 6 A. 7 8 9 10 11 12 13 14 Q. 15 A. 16 17 18 19 20 21 Q. 22 A. 23 Please discuss the Company's approach to calculating wind integration costs? The remainder of my testimony discusses the Company's treatment of wind integration costs. As part of the 2011 IRP, the Company performed an extensive Wind Study on the impact of integrating wind generation into its resource portfolio. The Wind Study was completed after reviewing the issues and concerns raised by various parties in Idaho and other jursdictions, such as whether the wid integration costs should be studied independent of load, the amount of additional reserves needed to integrate the wind generation and what resources should be utilized to serve the additional reserve requirements. Please describe the Company's Wind Study. The purose of the Wind Study is twofold. First, the Wind Study quantifies how wind generation affects the amount of additional reserves needed to maintain reliability. Second, the Wind Study determines the costs of integrating wid generation by measuring how system costs change with changes in operating reserve demand, and by measurg how system costs are affected by daily system balancing practices. What are the additional reserve requirements? The Wind Study identified additional regulation reserve requirements in two categories: regulating services that deal with load and wind varabilty in 10- REDACTED Duvall, Di- 16 Rocky Mountain Power 1 2 3 4 5 6 7 8 9 Q. 10 11 A. 12 13 14 15 16 Q. 17 A. 18 19 20 21 22 minute intervals, and load following servces that deal with load and wind variabilty over hourly time intervals. Both services respond to the up and down varations of wind generation. That is, the additional reserve requirements to integrate wind generation into the Company's resource portfolio consist of regulating up, regulating down, load following up and load following down. The Wind Study performed analyses of additional reserve requirements for load only (excluding wind generation) and for wid net of load (including wind generation), based on historicall0-minute data for the Company's system. In order to provide for this additional regulating and load following requirement, what changes in operations has the Company had to make? Given the size of the wind portfolio, and the possibilty of rapid variations in wind generation, the Company has had to commit its gas-fired generation units to be able to quickly respond to the magnitude of changes. At times, this "must-ru" operation requires gas-fired generation units to ru when it would otherwise be uneconomic to do so, thereby adding to the wind integration costs. How did the Company incorporate the results from the Wind Study? The amount of the total regulation reserve requirements to meet system variability is 337 average MW and 196 average MW for the east and west sides of the Company's system, respectively, for a total of 533 MW system-wide. The 533 MW is an increase of approximately 267 MW from the 2010 GRC, which only captued regulation reserve requirement for load variabilty, but did not include any incremental reserves for wind variability. REDACTED Duvall, Di- 17 Rocky Mountain Power 1 Q. 2 3 A. 4 5 6 7 8 9 10 Q. 11 12 13 A. 14 15 16 Q. 17 18 19 A. 20 21 22 Did the Company also change how it modeled its thermal resources in order to accommodate the additional reserve requirements in GRID? Yes. In order to accommodate the increase in regulation reserve requirements as identified in the Wind Study, the Company modeled the Curant Creek unit, and Gadsby units 4, 5 and 6 as must-ru units that are not subject to the logic of being committed to fUn only when economic. Modeling these resources as must-ru is consistent with the Wind Study, in which the Company concluded that it was appropriate in order for the model to be "reasonably aligned with actual operational characteristics of the east-side gas plants..."6 Does the Company believe that by reflecting the additional reserve requirements, instead of reflecting a dollar per MWh charge, it more accurately reflects the costs of integrating wind into its system? Yes. Allowing the GRID model to optiize the system, takig into consideration the additional reserves required to integrate the level of wind that is included in the GRID model, more accurately reflects the real-time operation of the system. Does this change in how the Company has modeled its wind integration costs address the Idaho Commission's concern that it is over recovering its costs associated with wind integration? Yes. The Wind Study is a verifiable study of the Company's costs associated with integrating wind generation into its systems. It clearly demonstrates the additional requirements for regulation reserves, and the requirement to have sufficient must- fUn thermal resources online to provide a quick response to the significant 6See Exhibit No. 36, the Wind Study, Page 25. REDACTED Duvall, Di- 18 Rocky Mountain Power 1 2 3 4 5 6 Q. 7 8 A. 9 10 Q. 11 12 A. 13 14 15 16 17 18 19 20 21 22 varations in wind generation. The level of reserves needed to meet variations in load and wind identified in the Wind Study are supported by actul reserves held by the Company for this purose in 2010. There is no reason to believe that the Company wil over recover its costs associated with wind integration. Customers are fuher protected by the Energy Cost Adjustment Mechanism ("ECAM"). Did the wind study also identify additional costs associated with day-ahead forecast errors for wind and load? Yes. Using the results of the Wind Study, the Company modeled $0.72/M for day-ahead forecast errors, or system balancing costs in the NPC study. Did the Company reasonably model the Wind Study results in GRI for the current fiing? Yes. The results from GRID reasonably reflect the impact of integrating wind generation into the Company's portfolio. The Wind Study addressed all the issues raised by various parties in Idao and other states, such as reserve requirements being modeled within GRID, the requirements for wind generation being considered net of load, studies supporting the impact of integrating generation from wind facilties, and the quality of data used to prepare the study. However, given the limitation of data inputs to the normalized studies, I believe that the GRID-modeled impact of integrating wind resources may understate the real costs. For example, the GRI model uses expected wind profiles of the wind projects which lack the variabilty reflected in the actual operations of the wind projects. REDACTED Duvall, Di- 19 Rocky Mountain Power 1 Q. 2 3 A. 4 5 6 7 8 9 Q. 10 11 A. 12 13 14 15 16 17 18 19 Q. 20 21 A. 22 23 Have you validated the Wind Study against the Company's actual experience? Yes. The Company computed the actual reserves carred durng 2010 to meet spinning and supplemental contingency reserves, and regulation reserves from recorded data. The actual regulation reserves carred on the system during 2010 was 629 MW. This is 96 MW higher than the 533 MW calculated in the Wind Study and modeled in GRID, thereby validating the Wind Study results as reasonable, if not low. Does the Company include system balancing costs for the non-owned wind projects and for projects located in the BPA's balancing area? No, the normalized NPC in the curent filing do not include system balancing costs for the wind projects located in the Company's balancing areas that the Company neither owns nor purchases the output. This is based on the assumption that the entities that own and/or operate those wind projects wil balance their own system prior to handing over their generation schedule to the Company. However, following that same logic, the normalized NPC include system balancing costs for projects located in BPA's balancing area: Leaning Juniper and Goodnoe Hils. Has the Company included wind integration costs for the non-owned wind projects located within its balancing area in this fiing? Yes. As discussed in the 2010 GRC, the Company is required to provide wind integration service to wholesale customers under federal law; the Company's Open Access Transmission Tariff (OATT) does not allow the Company to charge REDACTED Duvall, Di- 20 Rocky Mountain Power 1 2 3 4 5 Q. 6 A. 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 for this service. However, the Company wil file a wholesale transmission rate case with FERC on or before June 1, 2011, in which it wil propose a new Schedule 3A that, if approved by FERC, would allow the Company to charge these tyes of generators for regulation service in the futue. Does the Company believe that FERC wi approve its new Schedule 3A? Yes, although it is not known and measurable at this time. However, in Order No. 890-A, the FERC allowed transmission providers to propose to assess regulation charges to both generators sellng in the BAA and those sellng outside their BAA. This latter category are what are considered non-owned generation in GRID. The FERC stated that it would consider such proposals on a case-by-case basis. The FERC has approved transmission provider proposals to recover the costs of capacity associated with the provision of generator imbalance service though a proposed schedule seeking charges for providing generator regulation and frequency response service. Moreover, The Company's proposed Schedule 3A is consistent with the FERC's November 18, 2010, Notice of Proposed Rulemaking ("NOPR") on the integration of Variable Energy Resources ("VERs"), in which FERC proposed a generic rate schedule for the rates, terms, and conditions that apply to providing generator regulation and frequency response service. PacifiCorp's new Schedule 3A allows PacifiCorp to close the cost recovery gap with the tye of ancilar service that FERC has contemplated and accepted in the past for other public utilities. A transmission customer subject to Schedule 3A must either (1) take generator regulation and frequency response service from PacifiCorp or (2) REDACTED Duvall, Di- 21 Rocky Mountain Power 1 2 Q. 3 4 A. 5 6 7 8 9 10 11 Q. 12 13 A. 14 15 16 17 18 19 20 21 22 23 demonstrate that it has satisfied its regulation service obligation. Wil this new wholesale ancilary service charge, if approved by FERC, recover the costs of integrating wholesale customer wind projects? Yes. Regulation service is a tye of energy reserve service necessary for integrating resources into the transmission system. Regulation reserves account for the variabilty of load and generation on the transmission system on a moment-to-moment basis and over the course of an hour. The new Schedule 3A would apply to resources when they are exporting to load in other balancing authority areas and would result in additional revenues from non-owned resources that are not delivering power to serve the Company's native load. Do Idaho customers benefit from the Company providing Schedule 3A type services to the non-owned generation? Yes. As a balancing area authority, the Company owns and operates an extensive transmission network that it is required to operate safely and reliably for all of its customers, keeping all resources and loads in balance on a moment-to-moment basis. In addition, the Company is mandated to make its transmission network available to all generators in an open access and non-discriminatory fashion. By providing regulation and frequency response servce in addition to other transmission related services as a balancing authority, the Company ensures that its customers are served by a reliable system, with diverse resources. Moreover, any transmission revenues received from non-owned generation, which pays wheeling to the Company, are credited against retail revenue requirement and therefore have the effect of lowerig the cost of service for retail customers. If the REDACTED Duvall, Di- 22 Rocky Mountain Power 1 2 3 4 Q. 5 6 A. 7 8 9 10 11 12 13 14 15 16 17 18 19 20 FERC approves the new Schedule 3A, retail customers wil also receive additional revenues from providing servce to these wholesale wheeling customers. Do you have anything to add with regard to the issue of wind integration costs associated with wholesale transmission customers? Yes. I have been advised that because FERC has exclusive authority over the transmission and sale of electrcity in interstate commerce pursuant to the Federal Power Act, under the Supremacy Clause of the United States Constitution: a state utility commission setting retail rates must allow, as reasonable operating expenses, costs incured as a result of paying a FERC-determined wholesale price . . . Once FERC sets such a rate, a State may not conclude in setting retail rates that the FERC- approved wholesale rates are uneasonable. 7 Correspondingly, the Supremacy Clause would also require a state commission to allow as reasonable operating expenses costs that are incured as a result of operating consistent with a FERC-approved tariff. Based upon these . principles, I understand that because the Company is required by federal law to interconnect with wholesale transmission customers under the terms of the OATT, federal preemption precludes disallowing the associated costs, such as the costs regulation and frequency response services. 7Nantahala Power and Light Co. v. Thornburg, 476 U.S. 953, 956-966 (1986). REDACTED Duvall, Di - 23 Rocky Mountain Power 1 Q. 2 3 4 5 A. 6 7 Q. 8 A. Does the Company propose to update its filing in its rebuttal testimony for changes in net power costs, such as new contracts, fuel costs and the Offcial Forward Price Curve, irrespective of whether these changes increase or decrease net power costs? Yes. This ensures that the Commission has the most accurate and curent information available to it in setting rates for the test period. Does this conclude your direct testimony? Yes. REDACTED Duvall, Di - 24 Rocky Mountain Power ZOIJMAY27 AHfl:06 UTIL Case No. PAC-E-Il-12 Exhibit No. 35 Witness: Gregory N. Duvall BEFORE THE IDAHO PUBLIC UTILITIES COMMISSION ROCKY MOUNTAIN POWER Exhibit Accompanying Direct Testimony of Gregory N. Duvall Net Power Cost Report May 2011 ~( ) m ; o ;: ~ ~ 0 ¡g t l õ ' ~ g¡ Z ; : - . .. ? z s : Ç? " t ? g ti ; i ú ) : : (Q ( ) u i ¡ ¡ ~ r n ; ; = ï Z~ ( Q " t . . . C D 0 o ~ . . : E c ~ 0 g ¡ -: - gi u i Pa c l l i C o i p _l d a h o _ G R C _ ( C Y 2 0 1 1 L G O L D _ 2 0 1 1 _ 0 5 _ 0 2 Ne t P o w e r C o s t A n a l y s i s 12 m o n t h s e n d e d D e c m b e r 2 0 1 1 01 / 1 1 - 1 2 / 1 1 Ja n - 1 1 Fe b - 1 1 Ma r - 1 1 Ap r . l l Ma y . 1 1 Ju n - 1 1 Ju l - 1 1 Au g - l l 5e p - 1 1 OC t - 1 1 No v . 1 1 De . 1 1 Pu r c h a s e d P o w e r & N e t I n t e r c h a n g e Lo n g T a r m F i r m P u r c h a s e s AP S S u p p l e m e n t a l p 2 7 8 7 5 t, I 6 4 , 6 3 16 6 , 7 4 0 23 0 , 0 6 4 22 9 , 6 8 8 37 , 7 0 4 50 0 , 4 0 7 Bla n d i n g P u r c h a s e p 3 7 9 1 7 4 29 , 6 3 1 2, 5 1 7 2,2 7 3 2,5 1 7 2,4 3 5 2,5 1 7 2,4 3 5 2,5 1 7 2,5 1 7 2,4 3 5 2,5 1 7 2,4 3 5 2,5 1 7 SP A R e s e r v e P u r c h a s e 24 9 , 0 8 1 12 , 6 3 2 14 , 9 4 7 25 , 4 4 0 19 , 4 0 4 23 , 5 8 2 26 , 0 4 1 26 , 6 3 1 22 , 6 4 4 18 , 4 6 5 23 , 7 5 3 19 , 3 4 5 15 , 9 9 8 Ch e h a l i s S t a t i o n S e r v i c e 32 , 2 2 2 32 , 2 2 2 Co m b i n e H i l s W i n d p 1 6 0 5 9 5 4, 5 2 4 , 7 7 6 43 , 1 9 3 28 2 , 3 7 2 50 0 , 7 0 4 35 2 , 4 3 2 32 7 , 3 4 9 39 4 , 0 8 7 37 8 , 0 7 7 37 5 , 0 7 8 35 7 , 4 4 3 38 3 , 7 4 9 42 7 , 9 8 8 31 2 , 3 2 6 De s e r e t P u r c h a s e p 1 9 4 2 7 7 30 , 7 3 1 , 4 2 0 2,6 3 6 , 4 9 9 2, 4 4 , 8 4 8 2,3 2 7 , 2 6 0 2,2 6 4 , 4 7 8 2,0 4 1 , 2 5 4 2,1 5 9 , 8 4 2 2,7 5 2 , 7 8 7 2,7 8 0 , 6 9 1 2,7 2 4 , 8 8 3 2,7 6 6 , 7 3 9 2,7 9 4 , 6 4 3 2,8 3 6 , 4 9 9 Do u g l a s P U D S e t l e m e n t p 3 8 1 8 5 1,4 1 5 , 6 6 3 60 , 2 9 5 41 , 7 1 2 99 , 9 8 6 16 2 , 0 2 2 26 8 , 0 6 29 9 , 3 1 0 19 2 , 1 1 9 10 2 , 1 6 3 51 , 3 7 1 50 , 1 1 6 50 , 3 3 9 38 , 1 6 7 Ge m s t a t e p 9 9 4 9 2, 8 4 0 , 3 0 0 23 1 , 1 0 0 22 8 , 1 0 0 23 7 , 1 0 0 22 8 , 1 0 0 22 8 , 1 0 0 22 8 , 1 0 0 22 8 , 1 0 0 23 8 , 4 0 0 22 8 , 1 0 0 25 0 , 4 0 26 3 , 6 0 0 23 1 , 1 0 0 Ge o r g i a - P a c i f i c C a m a s 7, 4 0 8 , 2 2 6 62 9 , 1 9 2 58 8 , 3 0 62 9 , 1 9 2 60 8 , 8 9 5 62 9 , 1 9 2 60 8 , 8 8 5 62 9 , 1 9 2 62 9 , 1 9 2 60 , 8 9 5 62 9 , 1 9 2 60 8 , 8 9 5 62 9 , 1 9 2 Gra n t C o u n t y 1 0 a M W p 6 6 2 7 4 6, 3 6 2 , 0 2 6 50 8 , 6 0 7 39 9 , 9 5 9 43 9 , 3 0 8 49 4 , 9 1 2 56 0 , 6 7 0 60 6 , 3 6 1 69 7 , 4 9 6 72 6 , 7 4 6 54 , 9 1 0 47 0 , 6 6 5 40 , 5 2 1 50 6 , 8 7 3 He r m i s t o n P u r c h a s e p 9 9 5 6 3 79 , 7 5 0 , 7 4 4 8,8 5 6 , 5 7 8 7,6 7 9 , 0 5 6 4,7 2 2 , 4 7 4 4,7 2 2 , 4 7 4 4,7 2 2 , 4 7 4 4,7 2 2 , 4 7 4 5,3 8 7 , 1 1 5 8,0 8 8 , 1 0 8 7, 5 3 , 3 7 2 7,2 2 6 , 0 8 6 7,4 6 2 , 9 0 1 8,6 3 0 , 6 5 1 Hu r r i c a n e P u r c h a s e p 3 9 3 0 4 5 14 6 , 4 5 9 12 , 2 0 5 12 , 2 0 5 12 , 2 0 5 12 , 2 0 5 12 , 2 0 5 12 , 2 0 5 12 , 2 0 5 12 , 2 0 5 12 , 2 0 5 12 , 2 0 5 12 , 2 0 5 12 , 2 0 5 IP P P u r c h a s e 28 , 5 6 5 , 3 7 1 2,6 1 0 , 3 9 9 2,2 7 3 , 1 1 3 2,3 3 6 , 9 0 3 1, 6 3 1 , 7 0 0 2, 5 2 2 , 0 3 2,4 5 3 , 9 5 9 2,6 1 6 , 7 0 4 2, 5 9 8 , 2 9 1 1,8 6 5 , 3 6 5 3, 0 4 3 , 9 0 3 2,1 8 1 , 9 8 1 2, 4 3 1 , 0 2 Ke n n e c G e n e r a t i o n I n c e n i v e 3, 8 0 , 4 4 99 , 9 0 8 1,3 8 8 , 5 7 8 1,4 5 5 , 8 4 2 47 6 , 4 9 0 38 8 , 6 3 0 LA D W P p 4 9 1 3 0 - 4 77 4 , 3 8 38 7 , 1 9 0 38 7 , 1 9 0 Ma g C o i p p 2 2 9 8 6 Ma g C o r p R e s e r v e s p 6 1 0 3 7 8 4,9 0 8 , 2 4 0 40 9 , 0 2 0 40 9 , 0 2 0 40 9 , 0 2 0 40 9 , 0 2 0 40 9 , 0 2 0 40 9 , 0 2 0 40 9 , 0 2 0 40 9 , 0 2 0 40 9 , 0 2 0 40 9 , 0 2 0 40 , 0 2 0 40 9 , 0 2 0 Mo r g a n S t a n l e y p 2 7 2 1 5 3 - 6 - 8 1, 5 3 0 , 0 0 0 51 0 , 0 0 0 51 0 , 0 0 0 51 0 , 0 0 Mo r g a n S t a n l e y p 2 7 2 1 5 4 7 1, 6 2 0 , 0 0 0 54 0 , 0 0 0 54 0 , 0 0 54 0 , 0 0 0 Nu c o r p 3 4 6 8 5 6 4,9 9 8 , 0 0 0 41 6 , 5 0 0 41 6 , 5 0 0 41 6 , 5 0 41 6 , 5 0 0 41 6 , 5 0 0 41 6 , 5 0 0 41 6 , 5 0 0 41 6 , 5 0 0 41 6 , 5 0 0 41 8 , 5 0 0 41 6 , 5 0 0 41 6 , 5 0 0 P4 P r o d u c t i o n p I 3 7 2 1 5 / p l 4 5 2 5 8 17 , 0 8 6 , 7 5 0 1,4 3 6 , 7 2 3 1,4 5 6 , 1 4 5 1,4 1 5 , 0 3 7 1,4 5 1 , 8 0 5 1,4 1 5 , 8 8 0 1,4 1 5 , 6 8 0 1,4 1 5 , 8 8 0 1,4 1 5 , 8 8 0 1,4 1 5 , 8 8 0 1,4 1 5 , 8 8 0 1,4 1 5 , 8 8 0 1,4 1 5 , 8 8 0 PG E C o v e p 8 3 9 8 4 12 0 , 0 0 0 10 , 0 0 0 10 , 0 0 0 10 , 0 0 0 10 , 0 0 0 10 , 0 0 0 10 , 0 0 10 , 0 0 10 , 0 0 0 10 , 0 0 0 10 , 0 0 0 10 , 0 0 0 10 , 0 0 0 Ro c R i v e r W i n d p I 00 3 7 1 5,0 4 1 , 6 8 8 61 4 , 8 1 4 48 5 , 4 5 49 0 , 6 8 7 38 4 , 2 9 5 36 7 , 6 6 0 27 7 , 4 7 0 19 7 , 8 1 2 23 9 , 0 5 8 31 0 , 5 1 5 44 , 0 4 1 60 5 , 4 7 8 62 , 4 0 8 Ro s e b u r g F o r e s P r o d u c t s p 3 1 2 2 9 2 5,0 9 1 , 0 4 5 74 0 , 8 7 3 67 4 , 1 0 7 74 7 , 8 2 1 72 3 , 2 5 0 74 0 , 8 7 2 72 3 , 2 5 0 74 0 , 8 7 3 Sm a l l P u r c h a s e s e a s t 70 3 , 1 2 2 84 , 9 4 4 69 , 2 7 2 63 , 4 4 57 , 7 1 5 50 , 6 0 3 46 , 1 0 2 47 , 3 2 7 45 , 7 0 8 47 , 1 5 9 52 , 2 8 0 57 , 7 9 6 80 , 7 6 8 Sm a l l P u r c h a s e s w e s 18 , 8 2 2 62 4 81 0 34 1, 8 8 5 6, 3 8 9 1, 3 3 8 1, 1 7 8 1, 6 1 7 1, 3 7 7 1, 3 2 5 97 7 94 6 Th r e e B u t e s W i n d p 4 5 7 20 , 5 9 8 , 4 9 7 2,3 0 6 , 6 5 0 1,5 9 7 , 9 1 3 2, 3 4 9 , 6 9 1 1,6 9 2 , 9 4 7 1,7 1 4 , 4 3 8 1,1 8 2 , 9 8 7 1, 0 5 4 , 4 4 0 1, 0 8 0 , 2 0 4 1, 4 2 2 , 0 7 3 1, 7 8 6 , 0 7 2 2, 0 0 6 , 0 3 9 2, 4 0 5 , 0 4 3 To p o f t h e W o r l d W i n d p 5 2 2 8 0 7 40 , 2 4 4 , 9 2 8 5, 2 9 4 , 2 9 2 3, 9 9 3 , 6 7 9 3, 8 0 7 , 4 4 5 3,0 9 7 , 1 2 1 2, 6 6 4 , 3 9 0 2, 4 1 9 , 1 3 8 1, 9 3 0 , 6 2 9 2,0 8 5 , 8 9 7 2,2 6 0 , 7 0 1 2,8 9 4 , 5 4 0 4,2 3 5 , 1 2 0 5, 5 6 1 , 9 7 9 Tr t - S t a t e P u r c h a s e p 2 7 0 5 7 9,5 1 5 , 7 5 8 78 8 , 4 7 9 79 2 , 9 2 66 2 , 7 0 3 66 4 , 7 1 1 74 5 , 7 6 2 77 9 , 4 1 3 83 7 , 5 2 5 86 , 2 3 4 83 , 4 3 6 86 7 , 1 8 9 83 3 , 9 9 5 84 9 , 3 9 0 Wo l v e r i n e C r e e k W i n d p 2 4 4 2 0 9, 8 4 , 2 4 5 72 9 , 8 6 0 57 5 , 5 4 1, 1 4 7 , 1 7 3 1, 1 0 3 , 8 8 5 1, 0 7 5 , 7 5 8 83 8 , 3 4 3 81 8 , 8 3 0 76 8 , 1 8 6 71 4 , 6 4 61 8 , 5 3 1 80 9 , 1 0 1 84 , 5 8 7 lo n g T e r m F i r m P u r c h a s e s T o t a l 28 9 , 1 2 5 , 4 4 5 29 , 2 2 4 , 9 5 6 24 , 4 2 9 , 2 4 9 22 , 8 5 2 , 9 5 6 20 , 5 1 2 , 2 0 20 , 9 5 , 7 1 3 21 , 1 8 3 , 0 5 8 23 , 8 5 8 , 5 8 8 26 , 0 3 , 2 5 6 22 , 2 6 0 , 2 3 7 24 , 1 6 4 , 3 1 3 25 , 0 8 6 , 4 4 1 28 , 5 6 4 , 4 7 5 Se a s o n a l P u r c h a s e d P o w e r Se a s o n l P u r c h a s e P o w e r T o t a l :: ( " m A l ;: g ¡ ~ 0 ff C D õ ' ~ '" Z _ . - . '" - .. 9 Z ~ Gl - 0 9 e ãl " . w : : (Q ( " ( J ¡ ¡ ~ m - 0 : ; ' ~~ . \ J ;Z . . C D 0 O~ i ' ~ ei ' O " .. - !l ( J Pa c l f l C o r p _l d a h o _ G R C _ C C V 2 0 1 1 L G O L D _ 2 0 1 1 _ 0 5 _ 0 2 Ne t P o w e r C o s t A n a l y s i s 12 m o n l h s e n d o d D e e m b e r 2 0 1 1 01 / 1 1 . 1 2 1 1 1 Ja n . l l Fe b - I I Ma r . l l Ap r . l l Ma y . l 1 Ju n . l l Ju l . l 1 Au g - I I Sø l l OC I . l 1 No v . l l De . l l Qu a l i f y i n g F a c i l i t i e s OF C a l i f o r n i a 4,0 7 8 , 9 5 6 36 3 , 6 8 8 47 8 , 3 6 62 8 , 0 9 2 71 7 , 9 0 8 72 1 , 2 8 0 52 8 , 6 7 7 15 3 , 2 0 4 74 , 8 2 3 62 , 8 4 59 , 4 6 8 89 , 5 4 6 20 1 , 0 6 3 OF Id a h o 4,6 8 2 , 8 2 6 30 2 , 9 1 5 27 8 , 0 5 2 34 8 , 6 4 5 38 6 , 8 4 4 50 1 , 8 2 7 57 9 , 3 1 2 49 0 , 2 4 9 38 9 , 6 3 1 34 , 0 9 3 37 1 , 4 3 8 35 4 , 6 6 2 33 4 , 1 5 8 OF Or e g n 18 , 5 4 1 , 1 1 8 1,6 1 7 , 2 6 2 1,6 0 9 , 5 5 3 1, 8 1 5 , 7 8 9 1,9 6 1 , 5 5 5 1, 9 5 6 , 1 7 5 1,6 8 4 , 3 1 4 1, 3 9 7 , 0 5 9 1, 3 0 1 , 2 8 6 1, 3 0 4 , 7 0 4 1, 1 9 3 , 9 7 2 1, 2 2 5 , 1 3 8 1, 4 9 4 , 3 1 0 OF Ut a h 1,2 1 1 , 4 7 5 93 , 4 1 2 89 , 3 0 0 95 , 6 9 3 10 6 , 4 8 7 10 8 , 5 3 11 2 , 8 4 0 10 4 , 9 4 6 99 , 2 2 5 10 0 , 3 4 10 6 , 9 0 9 10 6 , 1 4 6 97 , 6 3 9 OF W a s h i n g t o n 2, 9 9 1 , 8 6 1 19 5 , 1 0 7 19 7 , 4 0 8 20 8 , 7 8 1 22 0 , 8 3 6 27 5 , 5 8 2 29 , 8 8 9 32 9 , 7 9 1 31 7 , 8 8 1 29 7 , 8 5 4 24 2 , 5 2 4 20 7 , 8 5 5 20 1 , 3 5 2 OF Wy o m i n g 75 1 , 8 7 4 14 , 7 8 1 15 , 2 6 6 14 , 7 8 5 39 , 6 8 9 11 3 , 2 3 9 11 5 , 4 5 2 12 3 , 6 3 6 12 3 , 4 1 1 11 0 , 4 9 5 49 , 7 7 4 15 , 9 9 1 15 , 3 5 5 Bio m a s s p 2 3 4 1 5 9 O F 26 , 6 2 9 , 4 5 0 2,3 0 3 , 5 1 7 2, 1 0 6 , 3 9 9 2,3 0 3 , 5 1 7 2, 0 7 5 , 4 4 5 2,0 7 5 , 4 3 8 2,0 7 5 , 4 4 5 2,3 0 3 , 5 1 7 2,3 0 3 , 5 1 7 2,2 3 7 , 8 1 1 2,3 0 3 , 5 1 7 2,2 3 7 , 8 1 1 2, 3 0 , 5 1 7 Ch e v r o n W i n d p 4 9 9 3 3 O F 2, 5 3 8 , 0 1 6 23 7 , 3 7 8 28 9 , 3 3 4 29 9 , 1 5 9 11 4 , 3 9 7 12 2 , 0 2 6 11 5 , 7 9 5 11 9 , 1 5 0 20 7 , 8 4 17 4 , 6 1 8 27 5 , 5 3 3 27 6 , 1 9 1 30 6 , 5 9 1 DC F P p 3 1 6 7 0 1 O F 26 , 1 3 3 1, 3 5 4 95 0 1,3 2 1 94 4 1, 1 5 5 63 9 90 8 1,6 9 1 2,8 2 5 6,1 9 1 5,4 7 2 2,6 8 2 Ev e r g r e e n B i o P o w e r p 3 5 1 0 3 0 O F 2, 5 8 8 , 5 1 9 22 6 , 7 2 3 18 4 , 4 1 3 20 6 , 4 8 3 21 3 , 0 9 5 21 8 , 7 2 4 19 2 , 9 2 7 20 4 , 0 3 6 27 0 , 4 0 2 25 6 , 5 5 0 28 8 , 9 9 5 19 2 , 3 1 5 13 3 , 8 5 7 Ex x o n M o I I p 2 5 5 0 4 2 O F 30 , 9 5 0 , 4 2 9 4,1 6 3 , 6 1 0 3, 8 4 2 , 7 4 7 3,3 8 7 , 0 6 1 1, 7 7 4 , 4 6 0 1, 4 5 0 , 4 5 1, 2 8 0 , 7 0 7 1,9 8 2 , 9 4 7 2,6 1 7 , 2 2 5 2, 0 6 , 2 6 2 2,2 3 7 , 5 6 3 2,4 4 0 , 9 8 1 3,9 1 2 , 5 1 3 Ke n n e c o t R e f i n e r y O F 1,7 2 5 , 4 3 8 15 2 , 4 6 9 13 4 , 2 5 4 14 2 , 9 7 6 12 5 , 2 2 9 10 9 , 9 6 0 10 1 , 5 7 7 17 5 , 7 0 1 17 7 , 1 1 8 14 2 , 4 7 4 14 7 , 5 5 6 15 2 , 8 7 0 16 3 , 2 5 3 Ke n n e c o t S m e l t e r O F 6, 5 3 1 , 3 7 1 57 1 , 4 6 4 57 5 , 5 4 47 5 , 9 9 1 49 1 , 1 5 8 32 2 , 8 9 4 40 8 , 9 9 9 66 3 , 8 0 1 63 8 , 4 1 6 52 2 , 6 9 7 61 0 , 2 9 3 61 9 , 6 6 8 63 0 , 4 4 3 Mo u n t a i n W i n d 1 p 3 6 7 7 2 1 O F 8, 4 3 6 , 0 4 2 1, 1 9 7 , 3 0 4 76 6 , 0 0 8 78 9 , 5 7 2 59 2 , 2 5 4 49 7 , 4 4 9 36 2 , 5 1 40 1 , 4 3 8 54 , 1 7 0 63 1 , 3 9 6 71 1 , 4 7 5 93 0 , 3 5 1 1,1 1 2 , 0 4 3 Mo u n t a i n W i n d 2 p 3 9 8 9 O F 12 , 2 1 4 , 6 0 4 1, 7 4 6 , 9 5 3 1, 0 7 2 , 0 2 2 1, 1 2 0 , 2 1 7 80 6 , 1 7 0 86 6 , 8 6 3 69 3 , 9 4 7 78 7 , 8 6 8 84 8 , 2 9 6 79 6 , 5 4 1 84 7 , 0 4 1 1,1 1 4 , 8 8 7 1,5 1 3 , 9 9 9 Or e g o n W i n d F a r m O F 10 , 1 4 5 , 4 1 6 58 5 , 0 9 7 84 7 , 4 1 1 82 9 , 7 7 1 1, 0 1 3 , 2 0 4 1,0 2 3 , 7 2 4 1, 1 9 0 , 0 8 0 1,2 0 4 , 6 5 3 93 2 , 5 6 7 75 1 , 6 1 4 76 5 , 2 7 8 89 1 , 2 1 3 31 0 , 8 1 4 Pio n e e r W i n d P a r k I O F 43 , 1 9 0 43 , 1 9 0 Po w e r C o u n t y N o r t h W i n d O F p 5 7 5 6 16 , 3 2 1 16 , 3 2 1 Po w e r C o u n t S o u t W i n d O F p 5 7 5 € 14 , 7 1 6 14 , 7 1 6 SF P h o s p h a t e s 4, 3 8 5 , 4 4 4 32 2 , 3 8 5 31 1 , 3 5 7 36 9 , 9 7 3 40 3 , 2 2 5 27 2 , 5 4 5 37 5 , 8 2 7 41 8 , 2 8 0 41 0 , 1 3 5 40 5 , 7 6 7 42 5 , 3 3 3 32 8 , 5 9 4 32 2 , 0 2 4 Sp a n i s h F o r k Win d 2 p 3 1 1 6 8 1 O F 2, 7 2 6 , 7 1 0 17 5 , 5 7 7 19 1 , 3 7 0 16 9 , 4 2 6 15 9 , 6 9 1 16 4 , 4 5 2 23 7 , 4 3 9 28 0 , 8 6 5 33 8 , 9 8 4 27 3 , 7 8 6 22 1 , 3 1 2 24 5 , 4 1 4 26 8 , 4 1 4 Su n n y s i d e p 8 3 9 9 7 / p 5 9 9 5 O F 25 , 7 3 3 , 8 6 9 2,3 1 5 , 6 5 7 2,1 9 4 , 1 4 9 2,2 7 4 , 0 1 6 1,3 3 7 , 0 2 4 1,9 4 6 , 1 8 7 2,2 7 5 , 2 9 7 2,3 1 6 , 4 3 4 2,3 2 0 , 5 5 2,2 3 , 2 5 7 1,9 4 2 , 6 3 7 2, 2 6 0 , 1 0 7 2,3 1 6 , 5 4 9 Te s o o OF 1, 7 6 5 , 5 7 2 15 6 , 9 1 4 13 7 , 5 4 14 5 , 0 8 9 12 8 , 5 5 2 11 2 , 5 5 3 10 1 , 8 7 0 18 0 , 1 2 5 18 1 , 4 0 1 14 6 , 4 4 8 15 0 , 9 1 1 15 5 , 1 3 9 16 9 , 0 2 6 Th r e e m i l e C a n y o n W i n d O F p 5 0 1 3 f 1,3 6 7 , 3 8 13 0 , 4 9 3 13 9 , 3 4 9 15 5 , 2 7 5 14 4 , 1 9 7 18 3 , 2 2 5 17 0 , 3 1 7 15 0 , 1 9 0 15 2 , 1 1 4 14 2 , 2 0 6 US M a g n e s i u m O F 7, 0 4 2 , 4 6 71 1 , 0 7 3 65 7 , 6 8 4 62 2 , 6 8 2 51 3 , 1 6 8 44 8 , 5 9 6 34 , 3 4 1 58 0 , 3 9 8 56 8 , 9 9 7 52 7 , 4 1 7 67 2 , 5 4 5 62 8 , 0 7 0 77 1 , 5 1 4 Qu a l i f y i n g F a c i l t i e s T o t a l 17 7 , 1 1 9 , 1 8 0 17 , 5 7 5 , 1 2 4 15 , 7 1 8 , 4 5 8 16 , 4 0 4 , 3 1 5 13 , 3 2 5 , 5 3 1 13 , 4 9 2 , 6 8 2 13 , 2 2 1 , 2 7 1 14 , 3 8 9 , 0 9 14 , 8 1 9 , 6 8 6 13 , 5 2 9 , 0 0 13 , 6 3 , 2 8 5 14 , 3 7 8 , 4 2 3 16 , 6 5 , 3 4 4 Mid - C o l u m b i a C o n t r a c t s Ca n a d i a n E n t i l l e m e n t p 6 0 8 2 8 Ch e l a n . R o c k y R e a c h p 6 0 7 3,6 1 0 , 8 1 6 36 1 , 0 8 2 36 1 , 0 9 2 36 1 , 0 8 2 36 1 , 0 8 2 36 1 , 0 8 2 36 1 , 0 8 2 36 1 , 0 8 2 36 1 , 0 8 2 36 1 , 0 8 2 36 1 , 0 8 2 Do u g l a s . W e l l s p 6 0 2 8 3,4 9 6 , 2 4 9 28 9 , 7 4 0 28 9 , 7 4 0 28 9 , 7 4 0 28 9 , 7 4 0 28 9 , 7 4 0 28 9 , 7 4 0 28 9 , 7 4 0 28 9 , 7 4 0 29 , 5 8 3 29 4 , 5 8 3 29 4 , 5 8 3 29 , 5 8 3 Gr a n D i s p l a c e m e n t p 2 7 0 2 9 4 9,5 2 9 , 4 5 9 94 5 , 5 3 9 89 1 , 4 5 6 92 0 , 4 7 1 1,2 0 3 , 6 0 5 1,2 4 5 , 0 4 2 1,0 3 7 , 0 1 5 1,2 1 7 , 4 3 0 1, 0 3 9 , 2 3 9 1, 0 2 9 , 8 6 1 Gr a n t R e a s n a b l e (8 , 4 4 2 , 4 1 7 ) (7 0 3 , 5 3 5 ) (7 0 3 , 5 3 5 ) (7 0 3 , 5 3 5 ) (7 0 3 , 5 3 5 ) (7 0 3 , 5 3 5 ) (7 0 3 , 5 3 5 ) (7 0 3 , 5 3 5 ) (7 0 3 , 5 3 ) (7 0 3 , 5 3 ) (7 0 3 , 5 3 ) (7 0 3 , 5 3 5 ) (7 0 3 , 5 3 5 ) Gr a n t S u r p l u s p 2 5 8 9 5 1 2, 0 4 , 5 8 5 17 0 , 7 1 5 17 0 , 7 1 5 17 0 , 7 1 5 17 0 , 7 1 5 17 0 , 7 1 5 17 0 , 7 1 5 17 0 , 7 1 5 17 0 , 7 1 5 17 0 , 7 1 5 17 0 , 7 1 5 17 0 , 7 1 5 17 0 , 7 1 5 Gr a n t . W a n s p u m p 6 0 8 2 5 - - - - - - - - - - - - - Mid - C o l u m b i a C o n t r a c t T o t a l 10 , 2 4 2 , 6 9 3 1,0 6 3 , 5 4 1 1,0 0 9 , 4 5 8 1,0 3 8 , 4 7 3 1,3 2 1 , 6 0 7 1, 3 6 3 , 0 4 1,1 5 5 , 0 1 7 1,3 3 5 , 4 3 2 1, 1 5 7 , 2 4 1 1, 1 5 2 , 5 0 6 12 2 , 8 4 5 (2 3 8 , 2 3 6 ) (2 3 8 , 2 3 ) To t a l L o n g T e r m F i r m P u r c h a s e s 47 6 , 4 8 7 , 3 1 7 47 , 8 6 3 , 6 2 1 41 , 1 5 7 , 1 6 6 40 , 2 9 5 , 7 4 4 35 , 1 5 9 , 3 4 1 35 , 8 1 0 , 4 4 0 35 , 5 5 9 , 3 4 6 39 , 5 6 3 , 1 1 6 42 , 0 1 1 , 1 6 3 36 , 9 4 1 , 7 4 7 37 , 9 1 7 , 4 2 3 39 , 2 2 , 6 2 8 44 , 9 8 1 , 5 8 2 ~ Q ~ a i ~ l ß ~ 0 '" z - , ~ '" . . ,. 0 z š : G) ~ ! , g ei ) l w : i CO ( ) ( J ¡ ¡ Om - 0 5 . -e i I I - 0 z. . c o 0 , T ( I : E 0" " w (I c: ~ 0 . . c: - ig ( J Pa c i f i C O r p _I d a h o _ G R C _ ( C Y 2 0 1 1 L G O L D _ 2 0 1 1 _ 0 5 _ 0 2 Ne t P o w e r C o s t A n a l y s i s 12 m o n t h s e n d e d D e c m b e r 2 0 1 1 01 / 1 1 - 1 2 1 1 1 Ja n - I I Fe b - l 1 Ma r - l 1 Ap r - I I Ma y - I I Ju n . l l Ju l - l l Au g - I I SS P - I I OC t - I I No v - i i De - I I St o r a g e & E x c h a n g e AP G I I C o i o c k u m s 1 9 1 6 9 0 AP S E x c h a n e p 5 8 1 1 8 1 s 5 8 1 1 9 Bl a c k H i l s C T s p 6 6 7 6 1,9 1 4 , 0 2 8 13 5 , 5 5 5 98 , 3 8 3 l0 0 , 3 2 9 71 , 3 8 9 91 , 9 1 0 14 1 , 7 1 6 88 , 3 1 8 87 , 7 3 0 41 4 , 5 9 4 24 7 , 8 8 5 18 1 , 4 2 6 25 4 , 8 1 1 BP A E x c h a n g e p 6 4 7 0 6 l p 6 4 8 8 8 BP A F C I I W i n d p 6 3 5 0 7 BP A F C I V W i n d p 7 9 2 7 BP A P e a k i n g p 5 9 8 2 0 38 , 4 1 0 , O O O 4,8 0 1 , 2 5 0 4,8 0 1 , 2 5 0 4,8 0 1 , 2 5 0 4, 8 0 1 , 2 5 0 4, 8 0 1 , 2 5 0 4, 8 0 1 , 2 5 0 4, 8 0 1 , 2 5 0 4, 8 0 1 , 2 5 0 BP A S o . I d a h o p 6 4 8 8 5 1 p 8 3 9 7 5 1 p 6 7 1 (9 , 8 8 4 ) (1 , 3 3 6 ) (6 , 1 3 7 ) (5 ) (2 , 4 0 6 ) Co w l i t z S w i f t p 8 5 7 8 7 EW E B F C I p 6 3 5 0 8 / p 6 3 5 1 0 PS C o E x c h a n g e p 3 4 0 3 2 5 4,5 0 0 , 0 0 0 37 5 , O O O 37 5 , 0 0 0 37 5 , 0 0 0 37 5 , 0 0 37 5 , 0 0 37 5 , 0 0 0 37 5 , O O O 37 5 , O O O 37 5 , O O O 37 5 , 0 0 0 37 5 , O Q O 37 5 , 0 0 PS C O F C I I p 6 3 6 2 1 s 6 3 3 6 1 Re d d i n g E x c h a n g e p 8 6 2 7 6 SC L S t a t e L i n e p 1 0 5 2 8 To t a l S t o r a g e & E x c h a n g e 44 , 8 1 4 , 1 4 3 5,3 1 1 , 8 0 5 5, 2 7 4 , 6 3 5, 2 7 6 , 5 7 9 5, 2 4 7 , 6 1 9 5, 2 6 6 , 8 2 4 5, 3 1 1 , 8 2 9 5, 2 6 4 , 5 6 3 5, 2 6 1 , 5 7 4 78 9 , 5 9 4 62 2 , 8 8 5 55 6 , 4 2 6 62 9 , 8 1 1 Sh o r t T e r m F i r m P u r c h a s CO B 1,9 0 6 , 0 5 1 45 9 , 2 5 2 75 7 , 4 3 9 68 9 , 3 6 0 Co l o r a d o Fo u r C o r n e r s 9,2 0 1 , 1 3 7 1,1 9 2 , 7 8 5 1,8 3 7 , 0 0 2 1, 9 6 0 , 5 7 0 44 0 , 8 0 0 49 8 , B O 44 0 , 8 0 0 49 8 , 8 0 0 45 2 , 4 0 0 46 , O O O 47 5 , 6 0 0 46 4 , 0 0 0 47 5 , 6 0 Id a h o 38 2 , 3 3 8 5,4 8 1 15 4 , 4 7 2 22 , 3 8 5 Me a d 41 6 , 8 0 5 23 , 8 8 5 25 6 , 2 1 0 13 6 , 7 1 0 Mid C o l u m b i a 36 , 0 9 5 , 5 5 1 5, 5 0 7 , 6 7 2 7, 0 6 , 9 4 3,8 3 5 , 4 6 6 5, 3 4 8 , 8 4 0 3,8 3 5 , 3 4 0 2, 9 2 2 , 8 4 1, 7 2 6 , 4 5 0 95 9 , 4 0 0 1, 7 0 4 , 0 0 0 1, 1 3 9 , 8 0 0 1,1 1 2 , 0 0 1, 1 3 9 , 8 0 0 Mo n a 3, 1 6 8 , 5 5 8 1,2 3 6 , 2 7 3 1,0 5 , 0 1 9 87 8 , 2 6 7 Pa l o V e r d e 8, 6 3 1 , 0 2 2 1,1 7 1 , 9 1 0 1, 1 2 7 , 3 9 2 1, 0 3 6 , 2 2 0 1, 7 0 6 , 2 0 0 2, 0 1 6 , 7 0 0 1, 1 1 3 , 4 0 0 45 9 , 2 0 0 Ut a h 1,0 0 8 , 5 8 7 12 8 , 8 4 4 55 7 , 2 4 5 32 2 , 4 9 8 Wa s h i n g t o n We S l M a i n 1,4 1 8 , 8 4 0 49 , 8 5 89 , 9 9 6 91 , 4 3 9 1, 1 8 7 , 5 5 0 Wy o m i n g 21 , 4 9 5 11 , 0 3 0 4,6 7 5 5,7 9 0 ST F E l e c r i c S w a p s (1 5 3 , 7 4 1 , 8 4 1 ) (2 1 , 6 2 0 , 1 0 8 ) (1 9 , 2 2 7 , 2 3 3 ) (2 6 , 6 n , 2 6 1 ) (1 3 , 6 0 2 , 8 5 7 ) (1 2 , 0 4 , 3 8 0 ) (1 2 , 4 4 6 , 8 8 0 ) (6 , 9 8 5 , 3 8 7 ) (6 , 2 1 3 , 7 2 3 ) (6 , 5 3 7 , 5 4 0 ) (1 0 , 0 6 0 , 1 3 6 ) (9 , n 2 , 3 9 1 ) (8 , 5 5 3 , 9 4 6 ) ST F I n d e x T r a d e s ~ ~ - - - - - - - - - - - To t S h o r t T e r m F i r m P u r c h a s e s (9 1 , 4 8 6 , 3 3 4 ) (1 1 , 8 3 5 , 8 2 6 ) (6 , 3 1 7 , 0 3 1 ) (1 7 , 4 9 8 , 5 5 7 ) (4 , 9 1 9 , 4 6 7 ) (5 , 8 8 3 , 5 4 0 ) (7 , 9 6 9 , 8 4 0 ) (4 , 7 6 0 , 1 3 7 ) (4 , 8 0 1 , 9 2 3 ) (4 , 3 6 9 , 5 4 0 ) (7 , 9 8 5 , 5 3 6 ) (8 , 1 9 6 , 3 9 1 ) (6 , 9 3 8 , 5 4 6 ) Sy s t e m B a l a n c i n g P u r c h a s e s CO B 17 , 4 9 4 , 9 8 3 22 9 , 1 8 5 48 5 , 3 9 7 2,2 4 9 , 4 6 7 1, 8 4 7 , 1 6 9 1, 6 3 7 , 7 2 7 1,6 9 4 , 2 8 5 3,7 2 8 , 5 9 3 98 1 , 3 3 1 45 9 , 5 1 1 1,9 7 3 , 6 4 0 1, 5 3 0 , 4 3 4 69 8 , 2 4 5 Fo u r C o m a r s 11 , 8 0 , 9 8 7 82 , 7 5 3 14 0 , 1 5 3 1, 1 1 5 , 9 9 8 1, 0 5 0 , 6 8 6 1, 0 7 6 , 4 5 3 1,1 0 0 , 0 6 5 2,0 1 0 , O O 9 1, 1 5 5 , 8 4 7 82 1 , 3 4 37 8 , 6 9 1 1, 2 6 0 , 3 0 0 1,6 1 0 , 6 8 7 Me a d 8, 0 3 2 , 4 2 4 2,8 2 0 , 8 2 3 1, 5 7 3 , 7 3 1 1, 2 4 8 , 6 5 2 37 , 7 5 5 44 , 7 6 7 22 , 0 8 0 54 2 , 5 2 7 72 7 , 4 1 9 12 7 , 0 5 9 2,4 2 8 88 5 , 1 8 3 Mid C o l u m b i a 78 , 8 7 6 , 3 1 5 9,6 7 0 , 4 n 6,1 1 5 , 5 9 7 6,9 8 9 , 3 5 8 5,1 7 2 , 5 8 5 3,9 4 0 , 2 0 6 3,8 8 8 , 8 0 7 17 , 2 9 3 , 3 2 0 13 , 4 0 3 , 7 8 4 4, 4 5 , 4 7 4 3,8 0 8 , 7 4 2 3,7 5 7 , 3 0 3 38 , 6 6 5 Mo n a 25 , 5 1 1 , 5 9 0 1, 3 2 8 , 2 7 9 45 8 , 2 8 7 2,0 0 8 , 9 6 2 2,8 9 2 , 3 8 0 2, 9 8 , 4 8 5 4,0 4 1 , 5 1 7 2,1 0 5 , 7 4 6 69 0 , 6 2 5 82 1 , 5 1 1 2,6 8 8 , 3 2 4 1, 9 8 6 , 5 7 9 3,4 9 9 , 8 9 5 Pa l o V e r d e 21 2 , 1 0 4 48 , 5 9 7 51 , 1 5 1 10 2 , 0 3 10 , 3 1 8 SP 1 5 Em e r g e n c P u r c h a s s 47 1 7 2 1l - - ~ 1L - - - - - - - To t a l S y s e m B a l a n c i n g P u r c h a s e s 14 1 , 9 n , 5 7 5 14 , 1 8 0 , 1 2 8 8, 8 0 , 3 1 6 13 , 7 1 4 , 4 7 3 11 , 0 4 0 , 3 2 7 9,7 0 6 , 3 6 0 10 , 7 4 6 , 7 5 4 25 , 6 8 0 , 1 9 5 16 , 9 5 9 , 0 0 6, 6 8 2 , 9 0 8,8 4 9 , 3 9 8 8, 5 3 7 , 0 4 7,0 7 6 , 6 7 4 To l a l P u r c h a s e d P o w e r & N e i l n i e r C h a l ' , 7 9 2 , 7 0 1 55 , 5 1 9 , 7 2 8 48 , 9 1 9 , 0 8 3 41 , 7 8 8 , 2 3 8 46 , 5 2 7 , 8 2 0 44 , 8 9 0 , 0 8 4 43 , 6 4 8 , 0 8 9 65 , 7 4 7 , 7 3 7 59 , 4 2 9 , 8 2 1 40 , 0 4 4 , 7 0 2 39 , 4 0 4 , 1 6 9 40 , 1 2 3 , 7 0 8 45 , 7 4 9 , 5 2 1 ~ & ' ~ 0 : : 9" ( J : : (\ ( \ õ ' ~ ~ Z ; : " c .. ~ Z š : ~" ' ~ g (\ ~ w : : (Q ( " u i ¡ ¡ o i " ' - - .: i ; I I : : Z ~ ( Q " ' , ~ ( \ 0 0' : . . ~ c: i \ o ~ c: - ~ u i Pa c l f l C o r p 12 m o n t h s e n d e d D e c e m b e r 2 0 1 1 Wh e e l i n g & U . o f F . E x p e n s e Fir m W h e e l i n g 8T F i r m & N o n ~ F i r m To t a l W h e e l i n g & U . o f F . E x p e n s e Co a l F u e l B u r n E x p e n s e Ca r t n Ch o l l a Co l s t r i p Cr a i g Da v e J o h n s t n Ha y d e n Hu n t e r Hu n t i n g t o n Jim B r i d g e r Na u g h t o n Ra m p L o s s Wy o d a k To t a l C o a l F u e l B u r n E x p e n s e Ga s F u e l B u r n E x p e n s e Ch e h a l i s Cu r r a n t C r e e k Ga d s b y Ga d s b y C T He r m i s t o n La e S i d e Lit e M o u n t a i n To t a l G a s F u e l B u r n Ga s P h y s i c a l Ga s Sw a p s Cla y B a s i n G a s S t o r a g e Pip e l i n e R e s e r v a t i o n F e e s . T o t a l G a s F u e l B u r n E x p e n s e Ot h e r G e n e r a t i o n Blu n d e l l Win d I n t e g r a t i o n C h a r g e To t a l O t h e r G e n e r a t i o n Ne t P o w e r C o s t Ne t P o w e r C o s t l N e t S y s t e m L o a e l RECEIVED ioii MAY 27 AM lf: 07 Case No. PAC-E-Il-12 Exhibit No. 36 Witness: Gregory N. Duvall;"l BEFORE THE IDAHO PUBLIC UTILITIES COMMISSION ROCKY MOUNTAI POWER Exhibit Accompanying Direct Testimony of Gregory N. Duvall 2010 Wind Integration Study May 2011 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 1 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A _RlCN ENERGY lIGS COMPA PacifiCorp 2010 Wind Integration Resource Study September 1, 2010 PAC I FiCORP Rocky Mountain Power Exhibit No. 36 Page 2 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MlDMIiRtII £NERGY !iNGS COMPA 2010 Wind Integration Resource Study 1. Executive Summary The purose of the 2010 Wind Integration Study (the "Study") is twofold. First, the Study quantifies how wind generation affects the amount of operating reserve needed to maintain historical levels of reliabilty. Second, the Study tabulates the cost of integrating wind generation by measurg how system costs change with changes in operatig reserve demand and by measurng how system costs are affected by daily system balancing practices. Based upon historical and simulated wind generation data and historical load data, the Study shows that operating reserve demand for both regulation reserve service and load following reserve service increases with higher wind penetration levels. For puroses of this Study, regulation reserve service refers to operating reserves required by variabilty in both load and wind over ten-minute time intervals and load following reserve servce refers to operating reserves required by both load and wind variabilty over hourly time intervals. Table 1 summarizes how operating reserve demand for both regulation and load following servces increases as wind penetration levels grow from approximately 425 MW to approximately 1,833 MW. Table 2 depicts the change in operating reserve demand that is incremental to a load only calculation of the same tyes of reserve service. Table 1. Annual average operating reserve demand by penetration scenario. Load Only 425MW 1372MW 1833MW Regulation Up 97 105 137 137 West Regulation Down 72 84 120 120 Load Following Up 101 114 139 141 Load Following Down 106 113 132 133 Regulation Up 138 140 201 231 East Regulation Down 107 110 185 222 Load Following Up 139 144 207 245 Load Following Down 144 147 198 237 Table 2. Annual average operating reserve demand incremental to the load only scenario. Load Only 425MW 1372MW 1833MW Regulation Up 0 7 39 39 West Regulation Down 0 12 48 48 Load Following Up 0 13 38 39 Load Following Down 0 7 26 27 Regulation Up 0 3 63 93 East Regulation Down 0 3 78 116 Load Following Up 0 4 68 106 Load Following Down 0 3 54 93 September 1, 2010 Page 1.of63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 3 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall ÁM~N eNeGYflltGS COPA The costs of integrating wind as calculated in this Study include costs associated with increased operating reserve demand as outlined above and the costs from daily system balancing practices. Both tyes of costs were calculated using the Planning and Risk model (PaR), which is a production cost simulation model configured with a detailed representation of PacifiCorp's system. For each wind penetration scenaro, a series of PaR simulations were completed to isolate each wind integration cost component by using a "with and without" approach. For instance, PaR was first used to calculate system costs without any incremental operating reserve demand and then again with the added incremental reserve demand. The change in system costs between the two PaR simulations drves the integration cost calculation. Table 3 summarzes the wind integration costs established in this Study alongside those costs calculated as part of the 2008 Integrated Resource Plan. Table 3. Wind integration costs per MWh of wind generated as compared to those in the 2008 IR. Study 200IRP 2010 Wind Integration Study 2010 Wind Integration Study Wind Capacity Penetration 2,734MW 1,372MW 1,833MW Tenor of Cost 2o-Year Levelized 3-Year Levelized 3-Year Levelized Interhour / System Balancing ($/MWh)$2.45 $0.82 $0.86 ReseNe ($/MWh)$7.51 $8.03 $8.85 Total Wind Integration ($/MWh)$9.96 $8.85 $9.70 As shown above, the Study fmds that operating reserve demand and the associated costs increase with wind capacity penetration. System balancing costs, drven by day-ahead forecast errors for wind and load, trend similarly as wind penetration increases from 1,372 MW to 1,833 MW; however, as expected, system balancing integration costs are much lower than integration costs for operating reserves. September 1,2010 Page 2 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 4 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MlnElltC ENeRGY HOlllGS COPA 2. Data Collection 2.1 Overview The calculation of Operating Reserve demand was based on load and production data over the 2007 to 2009 period (the "Initial Term"). Figue 1 shows that over this period, ten-minute interval data was not available for all wind resources included in the Study. Nonetheless, PacifiCorp chose to use this data because it represented the best base of observed data available within the company, it includes significant concurent load and wind generation data, and it includes year-on-year variabilty in weather and other variables affecting load and wind generation levels. Figure 1. Raw historical wind production and load data inventory. Plant name Foote Crek Stateline. Combine Hills Leaning Juniper Woli.nne Creek Marengo Goodnoe Hills Marengo II Mountain Wind I Spanish Fork "0 Mountain Wind IIi: :i Rolling Hills Glenrock Glenrock II Sei.n Mile Hill Sei.n Mile Hill ii High Plains McFadden Ridge I Thre Buttes Dunlap I Rock Rii.r Size, MW Timeline 70.2 60.9 19 79.8 99 99 39 99 20 99 28.5 99 111 50 81 201.5 Ie ;ilW,u¡rffil, ..~i0~j¡¡¡¡ = Internal fine resolution data (10-mln, 1-hour) .' = Data to be dei.lope by technical adi.sor * Capacity represents portion of the plant in PacifiCorp's control area. September 1, 2010 Page 3 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 5 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A M_!ltCN ENI¡ßGY liWNGS COPANY The data inventory summarized in Figue 1 contains as much real, observed, concurent data as possible, owing to the volatile and unpredictable natue of wind generation output as well as the many fine variations available in real load data that can be difficult to captue with simulated data. Nonetheless, the data set selected for the Study contains gaps, and as a result, PacifiCorp utilized the services of the Brattle Group, the technical advisor that assisted with this study, to simulate missing wind data pertining to the Initial Term. The simulation of wind data is discussed at length in its own section later in this report. 2.2 Historical Load and Load Forecast Data The historical load data for the East and West Balancing Authority Areas was collected for the Initial Term from the PacifiCorp PI systeml. These data were used for all the calculations involving historical load in the Study. The hourly day-ahead load forecasts were gathered from PacifiCorp's load forecast group, as were the day-ahead hourly load forecasts used to set up the generation system through the Initial Term period. 2.3 Historical Wind Generation and Wind Generation Forecast Data 2.3.1 Overview of the Wind Generation Data Used in the Analysis Ten-minute interval metered wind generation data were available for a subset of the wind sites as summarized in Figue 1. The wind output data were collected by PacifiCorp at each physical project location using the PI softare system. In addition to historical wind generation data, the Study required historical day-ahead wind forecasts, modeled day-ahead wind forecasts for simulated data, and the creation of an ideal wind profie. All of these data sets were needed to establish wind integration costs using PaR and are discussed in tum below. 2.3.2 Historical Wind Generation Data As shown in Figue 2, a cluster of PacifiCorp owned and contracted wind generation plants is located in Pacific Power's service area (PacifiCorp's West Balancing Authority Area) and another is located in the Rocky Mountain Power service area (PacifiCorp' s East Balancing Authority Area). It is worth noting that two wind sites, Wolverine Creek in Idaho, and Spanish Fork in Utah are part of the East Balancing Authority Area, but are geographically distant from both the western and the eastern clusters. i The PI system collects load and generation data and is supplied to PacifiCorp by OSISoft http://www.osisoft.comìsoftware-supportwhat-is-piíwhat is PI .aspx. September 1,2010 Page4of63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 6 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall AM_RIN ENliGY I!GS COMV enerating stations used in this study. The available historical ten-minute wind generation data were examined to produce some initial statistical diagnostics for each site and between sites. For each site, Table 4 shows: (l) number of 10-minute interval data observations available, (2) standard deviation of observed capacity factors, (3) the minimum capacity factor, and (4) the maximum capacity factor. Small negative capacity factor values (that show up as the miimum) in the data are the result of power consumption associated with routine operation of the wind projects even durng times when the project itself is not producing energy. Table 5 shows the correlation observed among aggregate hourly load and wind generation data in 2008. By and large, hourly changes in load and wind generation output, which drve operational planing, do not appear to be correlated. September 1, 2010 Page 50f63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 7 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MltEllAN ENERGY l!GS COPAN Table 4. Statistical properties of wind site capacity factor data. Plant Name Number of Observations Standard Deviation Goodnoe Leaning Juniper Combine Hills Stateline Marengo Wolverine Creek Spanish Fork Mountain Wind Foote Creek Seven Mile Hill McFadden Ridge High Plains Glenrock 83,520 157,824 157,824 157,824 79,776 157,824 74,736 66,096 157,824 52,704 11,952 15,84 50,256 32% 35% 38% 24% 33% 29% 29% 29% 300,Æ 31% 34% 21% 2go,Æ Min Max 0%100 OO,Æ 100% -3%100% -1%100% -11%100% -1%100% -4%87% OO,Æ 1000,Æ -2%100% OO,Æ 100% -1%100% OO,Æ 67% OO,Æ 100% Table 5. Hourly correlation of system wind and system load. 1 Overall IRolling 6 hour IRolling 12 Hour January -2.5%m-2.9%1 -3.4% February -2.8%-0.6%~-1.7% March -0.4%-1.4%1 -2.2% April -6.4%-3.5%1 -5.9% May -10.4%-3.0%1 -6.4% ~June -12.0%-9.2%1 -11.9% July -12.4%-12.3%~-14.2% August -9.1%-8.4%1 -9.8% September -6.5%o I -4.00,Æ-0.6%. 1October-3.5%-4.8%1 -6.7% ~November -7.5%-3.6%1 -4.4% December -2.00,Æ 0.3%1 -1.1% 2.3.3 Historical Day-ahead Wind Generation Forecasts Day-ahead wind forecasts were collected from daily historical fies maintained by PacifiCorp commercial operations. The fies contained day-ahead hour-by-hour wind generation forecasts for the wind projects operating durng the Initial Term. For those projects not operating durg the Initial Term, day-ahead forecasts were created using the daily volumetrc day-ahead forecast error from projects having complete data sets. As such, these data were used to bootstrap2 the 2 Bootstrpping is a common statistical method used to estimate data by extrpolating from existig data. September 1, 2010 Page 6of63 PACIFIC Rocky Mountain Power Exhibit No. 36 Page 8 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall AM_EIlIlGYllllNGSCOM_V daily day-ahead forecast volumetrc errors for the 1,372 MW and 1,833 MW scenaros, and the daily error (positive or negative) was applied to simulated wind generation data to create a modeled day-ahead forecast. The modeled day-ahead forecast maintained the same general hourly shape as the simulated wind generation data but was shifted vertically hour-by-hour on an equal percentage basis to keep the aggregate volumetrc error constant. 2.3.4 Ideal Shape Wind Generation In order to isolate wind integration costs from other system costs, a flat production profie is required for PaR modeling. This profile, deemed the ideal wind shape for puroses of the Study, treats all the energy produced by wind projects as monolithic blocks. Comporting with stadad trading products among forward energy markets in the Western Interconnect, the energy produced in each 16-hour daily block between hour ending seven and hour ending 22 was treated as a single block. Similarly, energy produced in the 8-hour block between hour ending 23 and hour ending six was treated as a single block. For each block, the total energy delivered from wind generation is averaged, thereby flattening the generation pattern. 2.4 Wind Generation Data Simulation The technical advisor assisted PacifiCorp in developing the Study methodology and in supplementing the historical wind generation data with simulated ten-minute interval wind generation data. This section sumarizes the methodology used to simulate wind generation data and provides sample data and graphics to ilustrate the details involved in each step of the process. The overall approach to simulatig wind generation data involved taking an historical data inventory; addressing data quality issues in the data inventory; identifying gaps requirg simulation; and fmding the best suited relationship between pairs of sites; and using that relationship to approximate the wind output for periods with missing historical observations. However, it is worth noting that for sites with no historical data, the necessary numerical relationships were estimated between relevant locations by using simulated wind data made available by the National Renewable Energy Laboratory (NREL). Additional detail on simulation procedures is available in Appendix A. 2.4.1 Categorization of Historical Wind Data to Determine Simulation Scope The historical wind data were classified into thee groups to determine the periods requiing simulation for each site. The three categories are defmed in tu below, and Figue 3 depicts how each site was categorized. (1) Fully Available-this category refers to sites for which output data are available for the entirety of the Initial Term. Specifically, these wind plants include: Leaning Juniper, Combine Hils, Stateline, Wolverie Creek, and Foote Creek. These plants sum to 425 MW of capacity. September 1, 2010 Page 7 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 9 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MIDERl~ ENIGY lING$ COM!' (2) Partially Missing-refers to sites for which output data are unavailable for a portion of the Initial Term. The wind plants that fall into this category are: Goodnoe Hils, Seven Mile Hil, Marengo, Spanish Fork, Mountain Wind, McFadden Ridge, High Plains, and Glenrock. One important featue of the parially missing data profies is that. the missing portions are always chronologically located at the begining of the time period-once a partially missing data profie begins, it contains no fuher data "holes". These plants sum to 848 MW of capacity. (3) Completely Missing-refers to wind projects, for which no output data are available for the 2007-2009 Initial Term. Those sites are: Dunlap I, Rock River, Rolling Hils, Three Buttes, and Top of the World. These plants sum to 560 MW of capacity. Fi ure 3. Cate -.avale Dat develop by Tecca Advso 2.4.2 Simulation Process The simulation process used in the Study evolved to become iterative in natue to ensure that simulated wind generation data used to establish operating reserve . demand was reasonably aligned to the operating reserve demand calculated using observed wind generation data. As such, different methods of error sampling and simulation techniques (multiple linear, Tobit; for example) were evaluated in this manner. Tables 6 ilustrates an example of how operating reserve demand calculated from observed and simulated data were used to evaluate different error sampling and re-addition methods used in this iterative process for the West Balancing Authority Area. September 1, 2010 Page 8 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 10 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall AM__ ENEflGYOOlllllGS COI'Y Table 6. Comparison of operating reserve demand calculated from actual wind generation plant data and simulated wind generation plant data estimated using a least squares regression and applying different scaling of errors added back into the raw prediction. Actual Wind Generation Data Load Foil owi ng Up 15.0 Load Following Down (19.1) Regulation 15.5 Test (Developed Wind Data) Error Scaling (%) Load Following Up10 9.950 10.675 11.7100 12.4 Load Following Down (13.0) (13.9) (14.2) (15.9) Regulation 11.1 12.3 14.3 17.1 Several simulation attempts ended with values above the feasible generation capacity range, or values beneath zero. Attempts to add the error term back into the prediction (a necessar simulation step) also faced significant hurdles in developing reasonable results. The highly variable ten-minute output led to error terms with ranges larger than the simulated values in many cases, which would also test the boundaries of either zero or maximum plant capacity delivered. Several processes were attempted to retu a sampled error estimation back to the modeled estimate, per proper regression, including sampling of trcated error distrbutions, medians of the error distrbutions, and various bins of errors sampled and added back to the regression estimate. Various combinations of these methods were put through the operating reserve demand estimation calculations to assess whether the results were reasonable. Ultimately, the Tobit simulation method (described in more detail in section A,4.3) and a 3-step smoothed median of the sampled errors proved to offer reasonably stable results. Ultimately, the iterative simulation process produced a simulation methodology comprised of several sequential steps: (1) estimate the Tobit regressions; (2) using the regression coeffcients, generate estimates of the mean output of the predicted variable3 (3) calculate the regression residuals; (4) randomly sample the residuals according to predefmed simulated output ranges; (5) apply a non-linear 3-step median smoother to the sampled residuals; (6) add the smoothed residual series to the predicted mean output. A more detailed description of each step appears in Appendix A, and the resulting regression coeffcients appear in Appendix B. 3 These are generally referred to in the literatue as "y hat" September 1, 2010 Page 9 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 11 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MlDmilI Ef!l!Y lIlOGS COPAN 3. Methodology 3.1 Method Overview This section of the Study presents the approach used to establish the enumeration of operating reserve demand and the method for calculating wind integration costs. Ten minute interval load and wind data is used to estimate the amount of operating reserve, both up and down, needed to manage fluctuations in load and fluctuations in wind within PacifiCorp's Balancing Authority Areas. The operating reserve discussed here is limited to spinning reserve and non-spinning reserve, which are needed for regulation, load following, and contingency reserve services. For puroses of this Study, regulation service refers to the operating reserve required to manage the varabilty of load and wind generation in ten minute periods, and load following service represents the operating reserve required to manage the variability as measured in hourly periods.4 Contingency reserve, although mentioned, is supplied in accordance with the North American Reliabilty Corporation (NERC) standards and remains unchanged by the wind generation contemplated in this Study. Therefore, the operating reserve quantities discussed herein are only pertinent to supplying the demands of regulation and load following services, which are assessed in for load, and load net wind scenarios. Once the amount of operating reserve is established for different levels of wind penetration, the cost of holding the reserve on PacifiCorp's system is calculated using PaR. In addition to using PaR for evaluating operating reserve cost, the PaR model is used to estimate wind integration cost associated with daily system balancing activities. These system balancing costs result from the unpredictable natue of wind generation on a day-ahead basis and can be characterized as system costs borne from committing generation resources against a forecast of load and wind generation and then dispatching generation resources under actual load and wind conditions. 3.2 Incremental Operating Reserve Demand A dense data set of ten-minute interval wind generation and system load drves the calculation of the marginal reserve requirement in two components: (1) regulation, which is developed using the ten-minute interval data, and (2) load following, which is calculated using the same data but estimated using hourly variability. The approach for calculating incremental operating reserve necessary to supply adequate capacity for regulation and load following at levels required to maintain curent control performance was based on merging curent operational practice with a surey of papers on wid integration, as well as advisory from the technical advisor.s The Initial Term load data is used as the baseline case (zero wind generation) in each scenario. Coincident wind data (as observed, plus that simulated by the technical advisor) were added in increasing levels of wind capacity penetration to gauge the change in operating reserve demand. For puroses of the Study, the regulation calculation compares observed ten-minute interval load and 4 PacifiCorp's definitions for reguation and load following are based on PacifiCorp's operational practice, and not intended to describe the operational practices or terminology used by other power suppliers or system operators.5 The extemal studies PacifiCorp has relied on can mostly be found on the Utility Wind Integration Group (UIG) website at the following link: http:í/www.uwig.org/opímpactsdocs.html September 1, 2010 Page 10 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 12 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MlOfC i_V HOI.GS COMPAV wind generation production to a ten minute interval estimate, and load following compares observed hourly averages to an average hourly forecast. 3.2.1 Regulation Operating Reserve Service Demand With no sub-hourly clearing or imbalance market, PacifiCorp must plan to meet sub-hourly load (and load net of wind) deviations with its own resources. This includes generating units on automatic generation control (AGC), demand side management (DSM), and the ramping of flexible generation units in real time operation, which requires that existing units be committed and then dispatched to provide operating reserve. Wind varabilty among ten-minute intervals can represent a quantity of generation required to ramp up or down to maintain system stability. Regulation servce demand for wind generation variability was considered first. To parse the ten-minute interval wind variabilty from the ensuing load following analysis, a persistence forecast of the rolling prior 60 minutes was used to analyze the variation of each ten minute interval. The actual wind generation in each ten minute interval was subtrcted from the rollng average of the prior six ten-minute intervals, and the standad deviation was computed for each monthly period. This approach follows the one used by the National Renewable Energy Laboratory (NREL) for its recent "Eastern Wind Integration and Transmission Study". 6 RegulationwindlOmin = Pcps2 (WindJ Where: PCPS2 = The percentile of a two-tailed distrbution equaling the Balancing Authority Area's CPS2 performance 7 Windi == the wind forecast error defined as (WindActuallOmin -WindlO-minjorecast) Windio-min-forecast = the rolling average of the wind generation in prior six ten-minute intervals, also referred to as a persistence forecast of the rolling prior 60 miutes WindActuaIlOmin = the observed wind generation for a given ten-minute interval The load varability and uncertinty was analyzed comparing the ten-minute actual load values to a line of intended schedule, which was represented by a line interpolated between an actual top- of-the-hour load value and the next hour's load forecast target at the bottom of that (next) hour. A sample of how the intended schedule compares to actul load data is shown in Figue 4. The method approximately mimics real time operations process for each hour. At the top of the given hour, the actual load is known and a forecast for the next hour was made. For the puroses of this study, a line joining the two points was made to represent the ideal path for the ramp or decline expected within the given hour. The resulting actual ten-minute load values were 6 NRL, Eastern Wind Ìntegration and Transmission Study, prepared by EnerNex Corporation, (Januar 10,2010), p.143. The report is available for download from the following hyperlink: http://vvvvw .nre1.gov/windlsystemsintegrationlpdfs!20 1 O!ewits fmal report.pdf 7 The Control Pedormance 2 is a reliability standard is maintained by the Nort American Electrc Reliability CounciL. A defmition is available on page 30f the document at the following hyperlink: http://www.nerc.comífies/Reliability Standards Complete Set 2010Jan25.pdf September 1, 2010 Page 11 of 63 . PACIFICORP 1141_ eNEGY HOGS COMPA Rocky Mountain Power Exhibit No. 36 Page 13 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall compared to this straight line so as to produce a strp of error terms, as depicted in Figue 5 with data from Februar 2009. The errors were assembled monthly and their Regulation demand estimated similarly to the method used for the 10-minute values of the wind data: RegulationloadlOm¡n == Pcps2 (LoadJ Where: Load¡ == the load forecast error, calculated similarly to Wind¡ 7000 Figure 4. Sample of intended schedule ten-minute load estimate and observed system load. 6500 ------------------ 6000 '"....II~5500II.llCI~ 5000 4500 -10 Minute Load Estimate ..;';Q;Aë:tüãICoad ,. 4000 ~~ ~w ~w ~w~w MW ~~ ~ w ~~ ~w~w MW" w"w"w"M NN Mm e e ~~ww ~ ~ ~~~~oo MMN N Mm e eMMMMMMMMMM Time September 1, 2010 Page 12 of 63 Rocky Mountain Power Exhibit No. 36 Page 14 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall lINGS COMPA Figure 5. Variabilty between the line of intended schedule and observed load with errors highlighted by green arrows. 6400 6300 6200 ¡ !6100 .. :æ 6000 5900 5800 w,.,,_a_~"" """"_",","' ___~"',.A'","'"~'~T'~~'.____,.,"'''"..,.,.,..... ......~w,.~_"'''".,_,_,'''''''.,-,-",.-----.,_, _u"'''____, _, , _,__y_.~~~_,,_vv m."""_.N'~Ah_'""___~.A_____~.w,.__""'"'_r'___""., .10Minute Load Estimate . il Actual Load ~ .II II.t i.~ ..II ,"I ¡I~ I , 2:00 2:10 2:20 2:30 2:40 2:50 3:00 3:10 3:20 3:30 3:40 3:50 4:00 TIme As the ten-minute load and wind errors each represent unpredictable change in the need for dispatchable generation, their variability was assessed separately and combined. The regulation demand of load "Ilet wind generation was estimated assuming short term variations in load are not correlated with changes in aggregate wind generation output though the use of a geometrc average (shown for Regulation Up): ReguiationUP10min = RegulationioaduplOmin 2 + RegulationWindupiomin 2 As the need for regulation service can vary whether the wind is up or down, both Regulation Up and Regulation Down services were estimated at each end of the error distrbutions. A sample of the errors logged for the same period, for load and wind, are shown in Figue 6. The independence of the forecast errors for wind and load was assumed. These errors, or differences between forecast and actual, comprised an estimate of the demand made on regulation service operating reserves durg power system operations. These differences were calculated for every ten minutes of operation through the Initial Term period, and separated into monthly bins for fuher analysis. September 1, 2010 Page 13 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 15 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A M1O£IIC EIlEmY llllGS OOPA Figure 6. Independent forecast errors in ten-minute interval load and wind generation (December 2008, approximately 890 MW of wind penetration). 150.00 I 100.00 3: :e 50.00..o....wt; 11u l! 0.00~ ~~,,-----.--..i-,,-- j! -50.00 ... -- -100.00 Time Analyzing the results on a monthly basis as opposed to grouping all the calculations together annually allowed for the fact that some months' power service actually required less regulation (for example, July and August) than others, and so costs could be more accurately attibuted with a weighted average of results as opposed to grouping the entire year's operations into a single analysis bin. This is due to operating reserve being employed to manage the tails of the distrbutions involved, and a single annual bin would apply the greatest tail occurences to the entire year, as opposed to only the month in which it occurs. Figue 7 demonstrates the resulting distrbutions of regulation demand for wind generation, where regulation down demand is the negative side of the distribution. The vertical lines drawn on Figue 7 ilustrate the operating reserve threshold defmed in the Study and data labels are added to denote outlying data points. Similarly, Figue 8 ilustrates the resulting distrbution of regulation demand for load, where regulation up demand is the positive side of the distrbution. September 1, 2010 Page 14of63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 16 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall AM__ .NEGYlIIIGS COMl' Figure 7. Wind Regulation errors plotted for the Mays ofthe Initial Term at the 1,372 MW wind ca aCI enetration leveL. 6000 Operating Reserve 5512 Threshold 5000 4000..-CQI...0.5 'õ 3000~QI.. E"Z 2000 595 110 1000 ¡ 1 1 2 3 04,8234 3 1 1 0 448 426 384 342 299 257 215 173 131 100 68 26 .16 .58 -97 -122 -164 -206 -248 .290 -332 -374 Figure 8. Load Regulation errors plotted for the Mays of the Initial Term. 10000 9000 8000 7000 VI..c 6000...,.u .E..50000..ai.i E 4000:iz 3000 2000 1000 10 ......... ........ ................ ............_.... .........8800. Operting Reserve Theshold 2 7101311111 22612182202218631703154 1385 1225 106 907 747 58 429 269 133 30 -114 -209 -368 -527 -687 -84 Megawatts September 1, 2010 Page 150f63 CIFI(QRP Rocky Mountain Power Exhibit No. 36 Page 17 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall 3.2.2 Load Following Operating Reserve Demand PacifiCorp maintains system balance by optimizing its operations to an hourly forecast with changes in generation and market activity. This planing interval represents hourly changes in generation which are assessed within roughly 20 minutes each hour to account for a bottom-of- the-hour (:30 after) scheduling deadline. Taking into account the conditions of the present and the expected load and wind generation, PacifiCorp must schedule generation to meet demands with an expectation of how much higher or lower system load (net of wind generation) may be. PacifiCorp's real-time desk updates the next hour's system load forecast fort minutes prior to each operating hour. This forecast is created by comparing the curent hour load to the load of a similar-load-shaped day. The hour-to-hour change in load from the similar day and hours (the load delta) was applied to the "curent" hour load and the sum is used as the forecast for the ensuing hour. For example, on a given Monday the PacifiCorp operator may be forecasting hour to hour changes in system load by referencing the hour to hour changes on the prior Monday, a similar-load-shaped day. If the hour to hour load change between the prior Monday's like hours was 5%, the operator wil use a 5% change in load as the next hour forecast. As for the corresponding short term operational wind forecast, the hourly wind forecast is done by persistence; applying the instantaneous sample of the wind generation output 20 minutes past the current hour to the next hour as a forecast and balancing the system to that point. The resulting operational modeling process therefore went as follows; at the top of the hour, wind generation output, dispatchable generation output, and load values were sumarized, and trended using the methods above. The result was compared to the next hour's schedule for gaps as soon as possible, with the generation and load values updated at roughly 20 minutes past the hour. In real time operations, this result would then be balanced through a combination of market transactions and scheduling adjustments to PacifiCorp resources to produce a balanced schedule for the ensuing hour; with all transactions having to be complete by 30 minutes past the hour. Meanwhile, for puroses of the calculation made in this Study, the hourly wind forecast consisted of the 20th minute output from the prior hour, and the load forecast was modeled per the approximation described above with a shaping factor calculated using the day from one week prior, and applying a prior Sunday to shape any NERC holiday schedules. Using the Initial Term data for PacifiCorp's Balancing Authority Areas, a comparson of the load and wind forecasts was implemented to measure the seasonal or annual trends in the varability between the hourly interval load and wind forecasts and the observed average hourly load and wind generation values. These differences were segmented into bins by load magnitude and wind generation magnitude using load and wind data, in order to facilitate making a weighted average of the reserves demand by load level and wind generation output leveL. An example of load and wind data segmented into bins appears in Figues 9 through 12. Figue 9 depicts forecast load in West Balancing Authority Area with a range of over and under predictions tied to Control Performance 2 (CPS2) performance leveL. Figue 10 shows the same data for the East Balancing Authority Area. In similar fashion, Figue 11 displays forecasted wind generation in September 1, 2010 Page 16 of 63 PACIFICO Rocky Mountain Power Exhibit No. 36 Page 18 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MitERtC ENeRGY HONGS COI' the West Balancing Authority Area with a range of over and under predictions consistent with a 97% CPS2 performance leveL. Figue 12 shows the same wind generation forecast data for the East Balancing Authority Area. Figure 9. Example of bin analysis for load following reserve service from load variabilty in the West Balancin Authori Area (Ma 2007-2009). 4000 3000 3500 ..Forecast Production a&AS&Under ..i ~500.. ~ 2000 1500 1000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Analysis Bin 7S00 7000 6500 -+Over '" ,-'" '" ,.... .....'" ,-.,,,. ""'" ,.... ,-'''. .",.. ,...w..........." .. ......"...,," ........ ...",.~.fgref.ClS.t,.~EQi:~~tign.,,"" ,_., w//~Under 6000 I rOO2;000 4500 4000 m. ....... ............ . ....m......... ..... ..mm...... ... .....mm..._.m.............. m. .... ..m. .... ...... ....... m. 3500 I 3000 . .1234567 L______.__.____.____________ 91011121314_~,iys~.iln 15 16 17 18 19 20 September 1, 2010 Page 17 of 63 Rocky Mountain Power Exhibit No. 36 Page 19 of 63 Case No. PAC-E-11-12 Witness: Gregory N. DuvallPACIFICORP A Mit_RI ENElG¥ i.NGS COMl' Figure 11. Example of bin analysis for load following reserve servce from wind variabilty at the 1,372 MW penetration level for the West Balancing Authority Area (May 2007- 2009). 600 ..........._...._. .-';UndË!"r....' I L:i 300iI I .. Forecast Production ~Over'" .~"__..~"_"_hhH~.~wh""._____.__"..¥_v._m,__,.__,,__,.,,_,,__.. __________________ 100 .... o 1 i 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 .100 Analysis Bin Figure 12. Example of bin analysis for load following reserve service from wind variabilty at the 1,372 MW enetration level for the East Balancin Authority Area (May 2007-2009). I :: 600 ..Under .. Forecast Production "R//.g- Dve r 500 :i j 1i ! !400¡ ..I..:: 300 200 100 o 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Analysis Bin September 1, 2010 Page 18 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 20 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A _!! E1lt'lG..llGS COI'V Probabilties implied by the population of each bin, representing the expected amount of time spent in each load state, were represented by the historical data. The percentile equivalent to the historical CPS2 performance of PacifiCorp was sampled above and below the median of each of the bins. The average CPS2 performance for PacifiCorp's East and West Balancing Authority Areas over the period 2004 to 2009 was just below 97%. As the goal of this Study is to incorporate wind integration in PacifiCorp's curent operations, the CPS2 performance of 97% was emphasized in these calculations. An assessment of the overall system power quality is a standalone topic that is beyond the scope of this Study, and thus, the Company assumed this level of reliabilty wil be maintained. The difference between the CPS2 percentiles and the median of the bins represents the implied incremental load following service for operating reserve demand within that bin. As each respective bin also has an implied probabilty. by the number of data points fallng within it, the volumetrc position over the study period was calculated as a simple weighted average. To fuher explain the calculation method for load following reserve demand, the followig example follows from the ilustration in Figue 10. To assess the load following up reserve. position for Bin 5, subtract the lower bound value (5,532 MW from the system load forecast of 5,687 MW to arrve at an estimate of 154 MW for the occurences within that bin. Integrating this process though all bins produced a composite load following up position for the East Balancing Authority Area in May, and the process was repeated for each month in the up and down directions. Wind generation was analyzed in exactly the same procedure, but with generation output representing the individual state variable. The wind and load reserve positions were combined using the root sum square calculation in each direction (up and down), assuming their variability in the short term is independent. ReserveSioadFollowing = LoadReserveSioadFoiiowing 2 + WindReserveSioadFoiiowing 2 3.3 Determination of Wind Integration Cost 3.3.1 Overview Owing to the varabilty and uncertinty of wind generation, each hour of power system operations featues a need to set aside increased operating reserve (both spining and non- spinning reserve), in addition to those set aside explicitly to cover load and contingency events which are inherent to the PacifiCorp system with or without wind. Additional costs are incured with daily system balancing practice that is influenced by the unpredictable natue of wind generation on a day-ahead basis. To derive how wind generation affects operating reserve costs and system balancing costs, the Study utilzes the PaR modeL. September 1, 2010 Page 19 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 21 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MItRl ENEllY l!NGS COMPA PacifiCorp's PaR model, developed and licensed by Venty Energy LLC, uses the PROSYM chronological unit commitment and dispatch production cost simulation engine and is confgued with a detailed representation of the PacifiCorp system. For this study, four different PaR simulations were developed for a range of wind penetration scenaros as defmed in Table 7. By carefully designig the four simulations, we were able to isolate wid integration costs associated with operating reserves and to separately calculate wind integration costs associated with system balancing practice. The former reflects integration cost that arises from short-term (within the hour and hour ahead) variability in wind generation and the latter reflects integration costs that arise from errors in forecasting load and wind generation on a day-ahead basis. Table 7. Wind penetration scenarios used in PaR, as a percentage of total fleet capacity. The four PaR simulations used for each penetration scenaro in the Study are summarized in Table 8. The first two simulations are used to tabulate operating reserve wind integration costs, while the third and fort simulations support the calculation of system balancing wind integration costs. Table 8 identifies how key input variables change among the simulations. The simulations were ru over the 2011 to 2013 forward term (three years), wherein 2007 wind generation and load data are used as inputs for 2011, 2008 wind generation and load data are used for 2012, and 2009 wind generation and load data are used for 2013. This calculation method combines the benefits of using actual system data available for the historic thee-year Initial Term period with curent forward price cures pertinent to setting the cost for wind integration service on a forward basis.8 PacifiCorp resources used in the simulations are based upon the 2008 IRP Update resource portfolio.9 8 The Study uses the March 31, 2010 offcial forward price cure. 9 Th~ 2008 Integrted Resource Update report, fied with the state utility commissions on March 31, 2010. The report is available for download from PacifiCorp's IR Web page using the following hyperlin: htt://v.i\vw.pacificorp.com!content/ darnpacificorp/doc/Energy SourceslIntegrated Resource Plan/2008IRUpdate/ PacifiCorp-2008IRPUpdate 3~31- i O.pdf September 1, 2010 Page 20 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 22 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A M10Al1N ENElGY HONGS COM_ Table 8. Wind integration cost simulations in PaR. 1 2011 - 2013 Actul Ideal Shape None None 2 2011 - 2013 Actul Actul Yes None Operating Reserve Integration Cost = System Cost from PaR simulation 2 less system costs from PaR simulation I 3 2011 - 2013 Day-ahead Day-ahead Forecast Yes NoneForecast Yes 4 2011 - 2013 Actual Actul Yes (Commitment from PaR Simulation 3 System Balancing Integration Cost = System Cost from PaR simulation 4 less system costs from PaR simulation 2 3.3.2 Calculating Operating Reserve Wind Integration Costs To assess the effects of varous levels of wind capacity added to the Balancing Authority Areas on operating reserve costs, each penetration scenario was simulated in PaR using both ideal (Simulation 1) and actual (Simulation 2) wind profies. Both the ideal and actual PaR simulations excluded System Balancing costs. The ideal wind profile is a "flattened" representation of the actual profie, where wind generation is averaged across on- and off-peak blocks. Such a profie requires no additional operating reserve to support wind generation variabilty, and as such, Simulation 1 only included an operating reserve needed for load variability. In summary, Simulation 1 included actual historical loads, ideal wind profies, and no incremental operating reserve to account for wind varability. Simulation 2 used the actual wind generation profiles, which reflect the 2007 to 2009 observed and developed Initial Term wind data as inputs for the 2011 to 2013 forward period. These actual wind generation profiles reflect the same variability used to derive the incremental operating reserve requirements needed to integrate wind generation. Thus, the second PaR simulation includes the incremental operating reserve demand created by the variable natue of wind generation as well as the actual, variable wind generation profiles. The system cost differences between these two simulations were divided by the total volume of wind generation in each penetrtion scenario to derive the wind integration costs associated with having to hold incremental operating reserve on a per unit of wind production basis. 3.3.3 Calculating System Balancing Wind Integration Costs PacifiCorp conducted another series of PaR simulations to estimate daily system balancing wind integration costs consistent with the wind penetration scenarios studied. In this phase of the September 1, 2010 Page 21 of 63 Rocky Mountain Power Exhibit No. 36 Page 23 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall HOUlNGS. COMPAV analysis, PacifiCorp generation assets were committed consistent with a day-ahead forecast of wind and load, but dispatched against actual wind and load. To simulate this operational behavior, two additional PaR simulations were necessary for each wid penetration scenario. Simulation 3 was used to determine the unit commitment state of generation assets given the day-ahead forecast of wind generation and load. Simulation 4 used the unit commitment state from Simulation 3, but dispatches units based on actual wind generation and load. This actual wind and load data is pulled from the Initial Term, and thus, is identical to the actual wind generation and load inputs used to derive operating reserve wind integrtion costs as described above. In both of these PaR simulations, the amount of incremental reserve required for each penetration scenaro was applied. The change in system costs between Simulation 4 and the system costs from Simulation 2 already produced in the estimation of operating reserve integration costs isolates the wind integration cost due to system balancing. Dividing the change in system costs by the volume of wind generation in each penetration scenario produced a system balancing integration costs on a per-unit of wind production basis. 3.3.4 Allocation of Operating Reserve Demand in PaR PaR Simulations 2 through 4 require operating reserve demand inputs that must be applied consistent with the ancilary services strctue native to the modeL. The PaR model distinguishes reserve tyes by the priority order for unit commitment scheduling, and optimizes them to minimize cost in response to demand changes and the quantity of reserve required on an hour-to- hour basis. The highest-priority reserve tyes are regulation up and rewlation down followed in order by spining, non-spinning, and finally, 30-minute non-spining. 0 Reserve requirements in the model need to be allocated into these PaR reserve categories and are expressed as a percentage of load. The regulation up and regulation down reserves in PaR are a tye of spining reserve that must be met before traditional spinning and non-spining reserve demands are satisfied. The incremental operating reserve demand needed to integrate wind generation was assigned in PaR as regulation up and regulation down. The traditional spinning and non-spining reserve inputs are used for contingency reserve requirements, which remain unchanged among all PaR simulations in the Study. The 30-minute non-spinning reserve is not applicable to PacifiCorp's system, and thus it is not used in this Study. 10 In PaR, spining reserve is defmed as unloaded generation which is synchronized, ready to serve additional demand and able to reach reserve amount within 10 miutes. Non-spining Reserve is defied as unloaded generation which is non-synchronized and able to reach required generation amount within 10 minutes. September 1, 2010 Page 22 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 24 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MW..II¡UIlAN ElIfßG¥ ll$ CO_V Note that given the hourly granularty in PaR, there is no distinction between operating reserve categorized as regulation and load-following in terms of how the model optimizes their use. Thus both regulation reserve service demand and load following reserve service demand are combined as a geometric average and input in PaR as regulation up and regulation down. Furer, owing to the hourly granularity of PaR and the fact that PaR optiizes dispatch for each distict hour, regulation reserves are effectively released for economic dispatch from one hour to the next. The PaR model requires separate inputs for spining operating reserve and non- spinning operating reserve. Table 9 sumarizes how the services for operating reserves are applied in PaR. Table 9. Allocation of operating reserve demand to regulation, spinning and non-spinning t . .PRllreserve ca ee:ories in a . Reserve Service PaR Regulation Up PaR Regulation Down PaR Spinning Reserves PaR Non-Spin Reservs RegulationUp,o..n RegulationUp,o"'n 0 0 0 RegulationDown,o"'n 0 RegulationDonlO..n 0 0 Load Following Up Load Following Up 0 0 0 Load Following Down 0 Load Following Dow .0 0 0.5*(5% of Hydro and Wind 0.5*(5% of Hydro and Wind Contingency 0 0 Generation output + 7% of Generation output + 7% of Thennal generation output)Thnnal generation output) Total Geometric Ai.rage of the aboi.Geometric Ai.rage of the aboi.Sum of the aboi.Sum of the aboi. 3.3.5 Satisfying Reserve Service Demand in PaR PacifiCorp's thermal and hydro units are able to meet the reserve demand entered in PaR as shown in Table 10. Regulation reserve is tyically held by units operating in automatic generation control (AGq mode. 11 Contingency Reserve is specified by the Nort American Energy Corporation in per http:íiwww.nerc.com/fies/BAL-STD-002-0.pdf . September 1, 2010 Page 23 of 63 PACIFICORP "MI1)AME1'N e_.. HONGS COMPA BEAR RIVE CARBNI CARBON 2 CHliS æOIL4 CLRWATER I & 2 COLSTRI 3 & 4 COPCOI&2 CRIG I &2 CUNT CR DA VB JOHNSTON I DA VB JOHNSTON 2 DA VB JOHNSTON 3 DA VB JOHNSTON 4 FISH CR GADSBYl GADSBY 2 GADSBY 3 GADSBY4 GADSBY 5 GADSBY 6 HAYDENl &2 HEISTON I HEISTON 2 HUER I HUER 2 HUER 3 HUINGlON I HUINGlON 2 JCBOYl JIM BRI I JIM BRI2 JIM BRI3 JIM BRI4 LAKESIDE LEOLO UTTlEMOUNAIN MERWIN MID-COLUMBIA NAUGHONl NAUGHON 2 NAUGHON 3 SWIF TOKEEESUDE WYODAK YAlE Rocky Mountain Power Exhibit No. 36 Page 25 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall September 1, 2010 Page 24 of 63 PAC I FICORP Rocky Mountain Power Exhibit No. 36 Page 26 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A _I'H £NElGY HOG$ COl' 3.3.6 Modeling gas plant utilization in PaR One of the objectives in calculating wind integration costs using PaR was to emulate observed real-time unit commitment and dispatch behavior of PacifiCorp's thermal plants durg the simulation period. A specific focus was placed on east-side gas plants capable of providing regulation reserve service. The commitment status of these gas plants, consisting of Curant Creek, Lake Side, and Gadsby units 4 through 6, was initially set to "must ru" in PaR to miror recent utilization of these units. In the PaR framework, must ru status means that the unit is commtted, but not necessarly fully dispatched, at all times. PacifiCorp then compared the resulting simulated capacity factors for the simulation year 2013 against actual plant capacity factors for 2009 keeping in mind that 2009 wind generation and load data are used as inputs for the 2013 PaR simulation year. Differences in the capacity factors were reasonably small. Given these fmdings, PacifiCorp concluded that PaR was reasonably aligned with actual operational characteristics of the east-side gas plants when setting Curent Creek and Gadsby units 4 though 6 as must ru. Consequently, this must run configuation was applied in PaR to circumvent the fact that PaR establishes unit commitment on price and not necessarily on operating reserve requirements. In this way, and consistent with recent operational practice, the Curent Creek and Gadsby units 4 though 6 are available for meeting operating reserve obligations even when out-of-the-money from a pure market dispatch perspective. The must ru setting on Curant Creek and Gadsby units 4 though six was applied in PaR Simulations 2 through 4. In each of these simulations, incremental operating reserve demand needed to integrate wind is applied in the model, and must-ru configuation ensures that the select set of east-side gas units wil be available to meet the added reserve obligation even at times when they are out-of-the-money. In contrast, PaR Simulation 1 does not include any incremental operating reserve demand, and thus, the must-ru setting was not used. 3.3.7 Transmission Topology in PaR PacifiCorp used the PaR transmission topology consistent with the 2008 IRP Update as shown in Figue 13. September 1, 2010 Page 25 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 27 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MllElIC e_GY HONG$ COMPAN Figure 13. 'l Load .. Generaion CD Purchase/Sae Mark ia ContracExhanges .. PacifiCorp TranslTsson -Owned I Firm Rights" .... PlannedEnergyGateayTransmission " 3.3.8 Carbon Dioxide Cost Assumptions in PaR Given the 2011 to 2013 forward term used in the Study, there was no C02 cost applied to fossil- fired thermal generating resources. This assumption simplifies any comparson of the calculated wind integration cost among the three forward simulation years and avoids the possibility of disparity between plant dispatch costs and wholesale electrcity market forward prices used over the term. This is in contrast to the 2008 IRP Update, in which PacifiCorp assumed that federal cap and trade carbon dioxide (C02) allowance prices go into effect in 2013, with prices starng at $8.58/ton in 2013 dollars and escalating at 1.8 percent per year thereafter. September 1, 2010 Page 26 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 28 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall AM~~l'fC ~foIlGY t!GSCOM_ 4. Results 4.1 Operating Reserve Demand Based upon historical and simulated wind generation data and historical load data, the Study shows that operating reserve demand for both reguation reserve service and load following reserve service increases with higher wind penetrtion levels. Table 11 sumarizes how operating reserve demand for both regulation and load following servces increases as wind penetration levels grow from approximately 425 MW to approximately 1,833 MW. Table 11. Annual average operatig reserve demand by penetration scenario. Load Only 425MW 1372MW 1833MW Regulation Up 97 105 137 137 West Regulation Down 72 84 120 120 Load Following Up 101 114 139 141 Load Following Down 106 113 132 133 Regulation Up 138 140 201 231 East Regulation Down 107 110 185 222 Load Following Up 139 144 207 245 Load Following Down 144 147 198 237 The increase in operating reserve necessar to support wind generation in grd operations is apparent in each of the penetration scenarios. For example, very little wind generation is added to the East Balancing Authority Area between the load-only and 425 MW scenarios, and understandably, there is little increase in the resultant incremental operating reserve demand. The same situation occurs between the 1,372 MWand 1,833 MW penetration scenarios on the West Balancing Authority Area, where again, there is little change to the calculated operating reserve demand. Additionally, as significant wind generation development impacts the East Balancing Authority Area between the 425 MW and 1,372 MW scenarios, and again between the 1,372 MW and 1,833 MW scenarios, there is clearly a proportionate growth of the operating reserve required to satisfy higher levels of wind penetration. Tabular monthly results for each Balancing Authority Area and for each tye of reserve service appear in Appendix C. For convenience, Figues 14 through 21 sumarize monthly operating reserve demand results. In reviewing these figues, it is helpful to compare the growth of estimated reserve demand per MW of wind penetration recognizing that most of the wind capacity in the 425 MW penetration scenario is in the West Balancing Authority Area and that most of the incremental wind capacity in the 1,372 and 1,833 MW penetrtion scenaros is in the East Balancing Authority Area. September 1, 2010 Page 27 of 63 PACIFICORP A M1DERlN ENERGY lIUlGS COPANY Rocky Mountain Power Exhibit No. 36 Page 29 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall Figure 14. Load following up operating reserve servce demand in the West Balancing Authority Area. 200 180 160 140 120 3: 100 :E 80 60 40 20 West Load Following Up 1 2 3 4 5 6 7 8 9 10 11 12 Month -Load -425MW ...-1372MW -1833MW Figure 15. Load following down operating reserve service demand in the West Balancing Authority Area. 200 180 160 140 120 ~ 100 80 60 40 20 West Load Following Down 1 2 3 4 5 6 7 8 9 10 11 12 Month -Load -425MW WØÆ71372MW -1833MW September 1, 2010 Page 28 of 63 ICORP llLONGS COMPAN Rocky Mountain Power Exhibit No. 36 Page 30 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall Figure 16. Regulation up operating reserve servce demand in the West Balancing Authority Area. West Regulation Up 180 160 140 3:120:E 100 80 60 1 2 3 4 5 6 7 8 9 10 11 12 Month -Load -425MW -1372MW -1833MW Figure 17. Regulation down operating reserve service demand in the West Balancing Authority Area. 160 140 120 100 ~ 80 60 40 20 West Regulation Down 1 2 3 4 5 6 7 8 9 10 11 12 Month -Load -425MW -1372MW -1833MW September 1, 2010 Page 29 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 31 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall " M_ERl i1NEIllY HOI.NllS COMPA Figure 18. Load following up operating reserve servce demand in the East Balancing Authority Area. East Load Following Up 350 300 ~250 -..._db",,~"'-R"'øø 3= 200 ,;#W'"'$',(-Load :E 150 ~-~..-425MW 100 -1372MW 50 -1833MW 1 2 3 4 5 6 7 8 9 10 11 12 Month Figure 19. Load following down operating reserve servce demand in the East Balancing Authority Area. East Load Following Down 350 300 250 100 50 -Load200 ~150 -425MW ~"'1372MW -1833MW 1 2 3 4 5 6 7 8 9 10 11 12 Month September 1, 2010 Page 30 of 63 CIFICORP Rocky Mountain Power Exhibit No. 36 Page 32 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall Figure 20. Regulation up operating reserve servce demand in the East Balancing Authority Area. East Regulation Up 350 300 250 2003:-Load ~150 -425MW 100 -1372MW 50 -1833MW 1 2 3 4 5 6 8 9 10 11 127 Month Figure 21. Regulation down operating reserve service demand in the East Balancing Authority Area. 200 ~ 150 100 East Regulation Down 300 -1833MW 250 -Load -425MW W.Øf_..1372MW 50 1 2 9 10 11 12345678 Month Figues 14 through 21 identify both the seasonal natue of the operating reserve required to cover wind integration services and the tendency for the services' demand to be increased in months where more wind energy is generated. The monthly variation in operating reserve demand is built into the costing of the services in PaR, considerig that the allocation of operating reserve for wind generation is less in the months where there is less need. September 1,2010 Page 31 of 63 PAC I FICORP A M_RlC¡' e_GY HOO1NGS COMPA 4.2 Wind Integration Costs Rocky Mountain Power Exhibit No. 36 Page 33 of 63 Case No. PAC-E-11-12 Witness: Gregory N, Duvall Tables 12 and 13 present the wind integration cost results for each wind penetration scenario. Costs are reported in both present value revenue requirement (PVRR) dollars and dollars per megawatt-hour of wind generation for each year in the study period. Levelized costs across the three year study term are also included in the far right column of each scenario table. Table 12. PaR simulation results for the load only scenario and the 425 MW wind penetration scenario. load Only I LevelizedTotal vanable cots 2011 2012 2013 Base (No Wind) Simulation 1 $ 1,192,79 $ 1,311,178 $ l,301,5n Simulation 2 N/A NIA NIA Simulation 3 1,188,903 1,30,920 1,286,758 Simulation 4 1,201,530 1,322,3n 1,313,055 calculation of Integraon Costs Operating Reserve (5im 2 less 5im 1) System Balancing (51m 4 less 5im 2) Total thousands $b$ . $ $ $ $ $thousands Wind Generaion (Actal) East Wind West Wind Total GWh bGWh Oprang Reserv System Balancing Total Wind Integraion b$ - $/MWh $ $ $/MWh $ $ $ $ $ $ $ .tlitl1b! I Levelized201120122013 1,141,30 $1,251,695 1,249,391 1,15,552 $1,261,783 $1,259,733 1,145,876 $1,251,190 1,241,733 1,152,34 $1,26,907 l,264,2n 9,244 $10,08 10,342 $ 25,830 1,796 $3,124 4,54 $8,09 11,04 $13,212 14,88 $ 33,924 534 60 520 1,44 754 79 665 1,937 1,28 1,396 1,185 3,3B3 $7.18 $7.22 8.73 $7.64 $1,39 $2.24 3.83 $2.39 $8.57 $9.46 1256 $10.03 September 1, 2010 Page 32 of 63 PAC I FICORP Rocky Mountain Power Exhibit No. 36 Page 34 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MltERIC ENElGV OONGS COPANY Table 13. PaR simulation results for the 1,372 MW and 1,833 MW wind penetration scenarios. 1400MW Total vañable cost 2011 2012 2013 I Løvollzod 1750MW 2011 2012 2013 I Lovollzod Base (No Wind)thousands $1,04,895 $1,141,572 $1,148,139 $1,014,831 $1,103,397 $1,112,343 $1,075,215 $1,172,782 $1,180,72 $1,053,713 $1,145,954 $1,156,774 $1,08,733 $1,179,114 $1,176,68 $1,06,866 $1,163,768 $1,163,482 1,077,117 1,175,126 $1,186,073 $1,057,087 1,149,48 1,162,164 thousands $28,320 $31,210 $32,58 $80,13 38,88 $42,557 $44,431 $109,512 $1,90 2,34 5,345 $8,165 3,374 3,530 5,390 10,60 thousands $30,222 33,554 37,934 $88,300 42,256 46,087 49,821 120,121 GWh 2,319 2,520 2,232 6,175 3,230 3,48 3,106 8,576 146 1,556 1332 380 1,46 1556 1332 3,805 GWh 3,781 4,076 3,564 9,98 4,692 5,04 4,438 12,38 $/MWh $7.49 $7.66 $9.14 8.03 8.29 8.44 1001 8.85 $0.50 $0.58 $L50 0.82 0.72 0.70 L21 0.86 $/Wh $7.99 $8.23 $10.64 8.85 9.01 9.14 lL23 9.70 Simulation 1 Simulation 2 Simulation 3 Simulation 4 Calculaon of Integraon Cost Operating Reseive (51m 21ess 5im 1) System Balancing (51m 41ess 5im 2) Total Wind Generaon (Actal) East Wind West Wind Total Operating Reserve System Balancing Total Wind Integraon The PaR model results demonstrate interesting trends in the component costs. Most notable is the reduction of system balancing costs for the 1,372 MW and 1,833 MW wind capacity penetratioIl scenarios when compared to the 425 MW wind capacity penetration scenario. This is due to the domination of load forecast error in the 425 MW scenario system balancing integration cost line item, where total system costs are divided by wind energy production to derive system costs on a per unit of wind generation basis. The system balancing costs stabilze as wind generation increases in the higher penetration scenarios. Additionally, the operating reserve integration costs increase with additional wind capacity penetration. The rate of increase in costs is outpacing the increased wind energy produced, resulting in a higher price per megawatt-hour of wind energy produced. Finally, it is noteworthy that the addition of wind generation capacity lowers overall system costs. Table 14 compares the results of the Study to integration costs developed for the 2008 IRP on a component by component basis using Levelized costs over the applicable terms. The primar differences in results are most apparent for inter-hour (2008 IRP)/system balancing (2010 Study) wind integration costs. This difference is explained by improvements in method. In the 2008 IRP, market transaction costs were used to estimate inter-hour integration costs, whereas the curent Study calculates system balancing integration costs derived from the operation of PacifiCorp resources. September 1, 2010 Page 33 of 63 PAC I FICORP Rocky Mountain Power Exhibit No. 36 Page 35 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MltM!E_ ¡N&V llL01NGS COMY Table 14. Wind integration cost comparison to the 2008 IRP. Study 20081RP 2010 Wind Integration StudyWind Capacity Penetration 2734 MW 1372 MW Tenor of Cost 2o-Year Levelized 3-Year levelized 2010 Wind Integration Study 1833MW 3-Year Levelized Expected to Day Ahead ($/MWh)$0.28 Day Ahead to Hour Ahead ($/MWh)$2.17 System Balancing ($/MWh)$0.82 $0.86 Subtotallnterhour / System Balancing $2.45 $0.82 $0.86 Intra Hour Reserves1 ($/MWh)$7.51 2010 Study Operating Reserves ($/MWh)$8.03 $8.85 Total Wind Integration $9.96 $8.SS $9.70 Assumptions Forward Price Curve Oct 2008, $8C02 Mar 2010, No CO2 Mar 2010, No CO2 1 - IRP resources were available to meet Operating Reserve demand before the in-service year, which lowers wind integration cost 4.3 Application of Wind Integration Costs in the 2011 Integrated Resource Plan The start of portfolio development for PacifiCorp's 2011 IRP is scheduled for September 2010. Portfolio development relies on the Company's capacity expansion optimization model, called System Optimizer. (Note that wid integration impacts are treated as an increased resource cost in the System Optimizer modeL.) The high-end wind capacity penetration scenario wil not be completed until after portfolio development is well underway. Until costs are assessed for the high-end wind capacity penetration scenario, PacifiCorp wil use the costs developed for the 1,833 MW penetrations scenario, totaling $9.70/MWh of wind generated power. September 1, 2010 Page 34 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 36 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MIPE!lN ENERGY HOLOlNG$ COMPAN Appendix A Simulation of Wind Generation Data A.l Detailed Discussion of Statistical Patterns of the Historical Wind Output Data From the available ten-minute interval historical wind generation data over the 2007 to 2009 Initial Term, there are four key observations. First, wind output has a seasonal pattern. Taking one plant as an example, Figue lA shows capacity factor data for Leaning Juniper in 2009. The red markers in the figue indicate the median of the distrbution, and the wide bar delineates the 25th to 75th percentiles of the distrbution. Figure lA shows the median, as well as the range of observed capacity factors in each month in 2009 for Leaning Juniper varies significantly. Second, the monthly standard deviations for capacity factor output are very different across sites in most months. Figue 2A compares the output patterns across June, July, and August of 2009 for Leaning Juniper and Combine Hils and shows that non-normality is evident in the data. Again, the red markers indicate the median of the distrbution, and the wide bar represents the 25th to 75th percentiles in the distrbution. Third, the commonly-accepted notion that wind output follows a pronounced diural pattern is only partially supported by the varous historical profies in the dataset, as apparent in Figue 3A. In general, such recurg patterns are more easily found in average aggregate representations of the data on hourly level, rather than by examining higher resolution ten-minute data. Figure lA. Leaning Juniper 2009 monthly capacity factors. September 1, 2010 Page 35 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 37 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MI1M.MlfRlAJI ENERGY HOGS COMPA Figure 2A. Comparison of Leaning Juniper and Combine Hils capacity factors. Figure 3A. Daily generation patterns of several PacifCorp wind plants. 100% 90% 80% 70%.. ~ 60%~ è 50%"" ou¡¡ 40%u 30% 20% 10% 0% Actual Capacity Factors ",., glenck -- --,. goodnoeom spaisbfor "'''-- stline "''''' leagiunper ' "'" mago ."." wolverek --Monthly Aver ;! ~ I ;¡ Finally, Figues 4A and 5A present the empircal distrbution of the 2009 capacity factor output of Leaning Juniper and Combine Hils, respectively. Both plants' hourly capacity factor data represent two key patterns to the study. One, that there are a very substantial number of zero generation hours for each station. Two, the output varies greatly through the potential capacity range of each generating station, implying the wind generation wil have the characteristic to vary from one time period to the next. This is different behavior than would be implied by a September 1, 2010 Page 36 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 38 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A M_RICJI ENERGY tIl.NGS COMPA strong bimodal diural pattern, which would imply very regular on/off behavior with and without wind. Fi ure 4A. Distribution of observed 2009 hourly ca A.2 Time Pattern of the Historical Wind Data The time-series properties of the wind generation data are also important to the Study. Initial data analysis revealed that the wind generation profies in the dataset were consistently September 1, 2010 Page 37 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 39 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A M_J\lC eNEVllOUllNGS COMPAN characterized by a slowly decaying auto correlation process, while their partial autocorrelations are significant up to 6 period lags. In other words, the wind data in a ten-minute period is heavily consistent with the previous 10-minute interval and, therefore, over time, the wind pattern could be described as influenced by its behavior in the previous time periods. Partial correlation measures the autocorrelation at a specific lagged time frame, while controllng for the effect of precedig lags. Partial autocorrelation is useful in determining the number of lagged terms to include as explanatory variables in a regression modeL. Figues 6A though 9A show the full and parial auto correlation factors for the Leaning Juniper and Combine Hils wind plants. Figues 6A and 7 A show that the predictive power fades regularly over time lag. Figues 8A and 9A show that the oscilating natue of wind generation is more apparent in the negative predictive power of the 2nd and 4th lags. Figure 6A. Autocorrelation coeffcients for successive ten minute lags in capacity factor for Leaning Juniper. September 1, 2010 Page 38 of 63 Rocky Mountain Power Exhibit No. 36 Page 40 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall HOUlGSCOM Figure 7 A. Autocorrelation coeffcients for successive ten minute lags in capacity factor for Combine Hils. Figure SA. Partial autocorrelation coeffcients for lags in capacity factor for Leaning Juniper. September 1,2010 Page 39 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 41 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A M_RIC EfIl!GY J.NGS COMPA Figure 9A. Partial autocorrelation coeffcients for lags in capacity factor for Combine Hils. A.3 Data Clean-up and Verification The source wind generation data were characterized by a number of issues that needed data clean-up, verification and, in some cases, adjustments. The first observed issue is that for certin records over various periods of time, the historical wind output data were zero. Those observations covered varying lengts of time and, in some instances, up to a few months. However, we noticed that the zero-value data blocks consistently occured only at the begining of a wind project's chronological energy output data and therefore it is suspected that those were probably periods when the plant had not yet been. fully commissioned. Thus, those observations are treated as "missing" and excluded them from the historical data set. Next, through our source data review, we identified that the output of certain plants seemed to have much smaller capacity factors and increased over time. This trend seemed to have extended beyond the natual volatility of wind generation for those wind sites and showed up as a gradual increase over time and reaching a maximum after a number of months. This observation seemed to suggest thatthe historical data were captuing the build-out of a wind site before it has reached its commercial operation date. As the maximum available capability though wind plant constrction on a daily basis was not documented, the decision was made to exclude wind output data for dates prior to the known commercial operation date for each wind site. As a result, the data set used for simulations was limited to include only date ranges that conform to the known commercial operation dates shown in Table lA. September 1, 2010 Page 40 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 42 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A M_RIC eNEMY HONGS COMI' Table lA. Summary of wind plant start dates and nameplate capàcity. Plant name Applied Commercial Operation Date Nominal Capacity (MW) Observed Max Output (MW) Dunlap I Goodnoe Hills Glenrock Glenrock III Rolling Hills High Plains McFadden Ridge I Leaning Juniper Marengo I Marengo II Seven Mile Hill I Seven Mile Hill II Combine Hills Wolverine Creek Mountain Wind I Mountain Wind II Three Buttes Top of the World Spanish Fork Foote Creek i Foote Creek II Foote Creek ill Foote Creek iV Rock River 11/1/2010 5/31/2008 1/17/2009 111 94 237 Data Unavailable 95 232 9/13/2009 99 148 10/10/2009 29 29 9/14/2006 101 103 6/26/2008 211 206 12/31/2008 119 123 6/17/2003 41 41 4/29/2005 65 65 9/29/2008 141 137 12/1/200 99 Data Unavailable 12/31/2010 202 Data Unavailable 7/31/2008 19 22 4/1/1999 95 137 The sites that were affected by these revisions were: · Goodnoe Hils (observations were set to missing for November 2007 through May 2008), · Marengo (observations were set to missing for February 2007 though May 2008), · Spanish Fork (observations were set to missing for April 2008 through JuI2008), · Mountain Wind (observations were set to missing for April 2008 through September 2008), · Seven Mile Hil (observation were set to missing for November 2008 though December 2008), · McFadden Ridge (observations were set to missing for June 2009 though September 2009), · High Plains (observations were set to missing for Februar 2009 through August 2009), · Glenrock (observations were set to missing for November 2008 though December 2008). September 1, 2010 Page 41 of 63 PAC1FiCORP Rocky Mountain Power Exhibit No. 36 Page 43 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall"M_RlAN .NEii lIGS COMPANY · That leaves five wind sites that were not affected by this adjustment -Leaning Juniper, Combine Hils, Stateline, Wolverine Creek, and Foote Creek. The second clean-up process involved understanding the aggregation of data and the interpretation of the plant size. The data provided to the technical advisor contained single wind output data stream for sites that share the same pricipal name but are distinguished as individual projects-those include Marengo and Marengo II, Mountain Wind and Mountain Wind II, Seven Mile Hil and Seven Mile Hil II, Glenrock and Glenrock III. The wind output data, which were collected on-site, did not distinguish between separate sharing the same name. The third clean-up involved the fact that the maximum output levels observed in the wind output data sometimes exceed the capacity officially available to PacifiCorp. The Study team decided to use the maximum output found in each wind profie data stream to be the de facto wind site megawatt capacity. We used this capacity level and converted each 1O-minute output into a capacity factor value ranging from 0 to 1.12 A.4 Wind Data Simulation Methodology A.4.1 General Description The overall methodology centered on using available data to estimate the missing data. To do so, the statistical relationships between pairs of sites were studied and those relationships were used to derive or estimate the wind output for periods that historical data are incomplete or missing. For example, if there was afully available set of historical data for site A, but partiallymissing for site B, the overlapping periods durng which historical data are available for both sites A and B were used to estimate the statistical relationship using that data. Then the technical advisor employed that statistical relationship and used the available data from site A for the period when site B has missing data to estimate wind data for that period. If site B has completely missing data, the technical advisor applied NRL's simulated data (from 2004-2007) to establish the statistical relationship between sites A and B and then applied that estimated relationship to the historical data of site A and again, estimated site B' s wind output accordingly. A.4.2 Wind Generation Estimation Model Specifcation In general, the modeling approach is based on the use of contemporaneously available ten- minute wind capacity factor data from fully available wind profies to simulate capacity factor data for profiles with partially or completely missing wind output. As prior figues demonstrated, ten-minute wind output exhibited a generally volatile profile with several notable featues. First, output from previous periods is highly indicative of the curent level of output, with the partial autocorrelations significant up to as many as six lags. Second, the diural patterns were harder to discern on a consistent basis. Given these characteristics and our preliminary analysis, we 12 The capacity factor represents the output at a given point in time as a frction of the maimum possible output for the wind project. For example, a capacity factor of 0.23 indicates that curent output is 23% of the total capacity of the wind site. September 1, 2010 Page 42 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 44 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall "M1OERICAN ENElGY lINGS COM_V chose to include six lagged terms in addition to the concurent wind output term in the model used to estimate the statistical relationship between pairs of sites. We have found that such specification allows us to captue the time-based behavior and time-dependence of the wind data used in the Study. This approach also captues some of the spatial relationship between the two sites-as wind moves from one site to the other, its impact on the other site is delayed in time. The equation below describes the general strctue of the model 13 : S.t A S.t B S. B S. B S. B S. B S. B S. BI ei = ao i ei + ai itei_i + a2 Itel_2 + a3 Itel_3 + a4 Itel_4 + a5 Itel_5 + a6 Itel_6 + 8 A.4.3 Wind Generation Estimation Model for Constrained Output An important challenge in specifying this model is the natue of the capacity factor variables. Capacity factor is used instead of absolute wind output levels to translate between small and large wind plants. By such a constrction, the wind output measured in capacity factor terms can only take values between 0 and 1 (or, equivalently 0% and 100%). Attempting to predict a limited dependent variable using a standard linear ordinar least squares (OLS) approach resulted in estimated values for the dependent variable (or sites with partially missing and completely missing historical data) that are outside the possible value range. For example, for given mean values of the explanatory varables, the linear OLS model might result in a predicted mean dependent variable value greater than a capacity factor of 100%. This is due to the fact that a linear OLS model does not limit the outcome range for the dependent variable. In the literatue, a model whose dependent variable is limited at either one or both upper and lower ends of its range is called a "censored" modeL. 14 A stadad approach for estimating a censored model is to use the Tobit regression modeL. The Tobit model was originally developed by James Tobin (1958)15 and employs an estimation technique, which recognizes the limited ("censored") range of possible values that the observed dependent variable can take. 16 As a result, predicted mean values for the dependent variable wil behave as expected and not exceed the natual capacity limits of 0 and 1, as specified in our case. The Tobit model uses a maximum likelihood process, which takes into account the probability of obtaining an observation that lies inside the censorig intervaL. In other words, Tobit tyically is used to estimate the likelihood of a value to be equal to some expected quantity. The model assumes that the tre value of the dependent variable (y*) is explained by a number of independent variables, where the regression error term (epsilon) is normally distrbuted with a zero mean. In addition, ify* is between 0 and 1 we observe y*, however, ify*.:O we observe 0 and, similarly, if y*::I, we observe 1. The maximum likelihood estimation uses the probabilty 13 We specify a regression model that has no constat term. 14 Greene, Wiliam H., "Econometrc Analysis", 5th Ed., Prentice Hall 2003, p. 764. 15 Gujarati, Damodar N., "Basic Econometrcs", McGraw Hil 2003, p. 616; Kennedy, Peter "A Guide to Econometrics," 5th Ed., MIT Press 2003, pp. 289-290. 16 Ibid. September 1, 2010 Page 43 of 63 PAC I FICORP Rocky Mountain Power Exhibit No. 36 Page 45 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A M_ilICAN ENElGV HO$ COMPAV of each individual observation being censored to estimate the regression coefficients.I7 In other words, the regression coeffcients are determined to ensure that their value maximizes the likelihood of obtaining the observed values of y* .18 In contrast to linear OLS regression, the Tobit regression model does not report an R-squared metrc, which tyically indicates the explanatory power of the regression model specification (with high R-squared value indicating stronger explanatory power). In other words, in the linear OLS regression, the adjusted R-squared measures the proportion of variance of the dependent varable that has been explained by the independent (right-hand-side) variables. There are a range of so-called "Pseudo R-Squared" metrcs that have been proposed in the literatue for use with maximum likelihood models, such as the Tobit modeL. However, their interpretation is not equivalent to the R-Squared in OLS. This is because estimates derived using a Tobit model are calculated via an iterative process designed to maximize the likelihood of obtaining the observations of the dependent varable, rather than to minimize variance.I9 The technical advisor used the statistical softare package STAT ArQ to perform the regressions using the Tobit modeL. The model specification uses the chosen explanatory variables and generates a censored prediction of y* where the relevant upper and lower censoring limits are taken into account.20 An example of the six-lag model the technical advisor settled upon for significance is below: Goodnoe/ = aoLeaningJuniperiB + aILeaningJunipe'"~1 + a2LeaningJunipe'"~2 + + a3LeaningJunipe'"~3 + a4LeaningJunipe'"~4 + a5LeaningJunipe'"~5 + a6LeaningJunipe'"~6 + 8 A.4.4 Using NREL's Wind Data to Facilitate Wind Simulation for Sites without Historical Information To simulate wind data of sites with no historical information, the technical advisor used the NREL wind data to estimate the statistical relationship between pairs of sites and then used the estimated relationship to simulate the necessary wind data. For sites with completely missing historical wind data, NREL sites are chosen to serve as a proxy wind profies. NREL's Western Wind Dataset was created by 3TIER for use in NREL's Western Wind and Solar Integration Study. The dataset was synthesized using numerical weather prediction (NP) 17 For example, see "STATA Base Reference Manual Release 11", Stata Corp. pp. 1939-1948; Maddala,G. S., "Limited-Dependent and Qualitative Variables in Econometrcs.", Cambridge University Press 1986, pp.159-162. 18 For more detailed description of the Tobit model, please see Maddala, G. S., "Limited-Dependent and Qualitative Varables in Econometrcs", Cambridge University Press 1986, pp.159-162. 19 For more information, please see: Long, J. Scott. "Regression Models for Categorical and Limted Dependent Variables" Thousand Oaks: Sage Publications, 1997; Freese, Jeremy and 1. Scott Long. "Regression Models for Categorical Dependent Varables Using Stata", College Station: Stata Press, 2006. 20 For more information, please see: Baum, Chrstopher F., "An Introduction to Modem Econometrcs Using Stata", College Station: Stata Press, 2006, p. 264. September 1, 2010 Page 44 of 63 Rocky Mountain Power Exhibit No. 36 Page 46 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A M10AMEllCiI ENfllGY fILOGS COMPAN models "to recreate the historical weather for the western u.s. for 2004, 2005, and 2006. The modeled data were temporally sampled every 10 minutes and spatially sampled every arc-minute (approximately 2 kilometers).,,21 We refer to this wind data set as the "NREL data". The first step in using the NRL Western Wind Dataset is to identify NREL-modeled sites that are the closest in geographical terms to the relevant PacifiCorp wind sites. These are called the "NRL proxies" for each corresponding PacifiCorp wind site. The technical advisor then estimated the statistical relationship between the pairs of NREL proxies (that correspond to PacifiCorp wind sites) and used the statistical relationship to carr out the rest of the simulation described above. PacifiCorp staff provided the technical advisor with the geographical coordinates (latitude and longitude) for the PacifiCorp wind sites as summarized in Table 2A. In addition, the NREL data contains comprehensive information on the geographical coordinates of all modeled sites.22 The technical advisor then determined the closest NREL proxy for each ofplant.23 Table 2A. NREL Proxies selected for pertinent PacifCorp plants. PacifiCorp Plant Name Closest NRL Site ID Distance (km) High Plains McFadden Rock River Rolling Hils Dunlap Three Buttes Top of the World 16676 0.516676 0.531422 0.423909 2.919280 0.823870 5.323803 4.8 Table 2A shows each PacifiCorp-NRL pair and the calculated distance between them. We should note that High Plains and McFadden Ridge share the same geographical location and, as a result, are paired with the same NREL-modeled site. As a result, High Plains and McFadden Ridge have identical simulated profies. (This is a fuction of the study's approach of simulating wind generation output based on geographical location rather than wind project name-for 21 htt://www.nre1.gov/windlintegrationdatasets/westernmethodology.html#methodology (accessed July 1,2010) 22 The main web portal for the NRL Westem Wind Dataset can be accessed at htt://wind.nre1.gov/Web nrel 23 Geographical coordinates for two points on the ear's surace can be converted to a straight-line distance using a range of alternative algorithms, which tae into consideration the shape of the ear and use trgonometric formulas to project and measure surace distances. For the puroses of this study, the Spherical Law of Cosines was used to calculate the distance between each relevant PacifiCorp wind site and every site in the Western Wind Dataset. Fore more information, please see: Weisstein, Eric W. "Spherical Trigonometr." From MathWorld -- A Wolfram Web Resource. http://mathworld.wolfrain.coinSphericaITrigonometr.html (accessed July 1,2010) Distace (km) = ArcCos( Sin(Latitude Pacificorp) * Sin(Latitude NRL) + Cos(Latitude Pacificorp) * Cos(Latitude NRL) * Cos(Longitude NRL - Longitude Pacificorp)) * 6371 km September 1, 2010 Page 45 of 63 PACIFiCORP Rocky Mountain Power Exhibit No. 36 Page 47 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MIOEl' EIIEIlY HOUlGS COMPA example, the same simulated profie is also used to represent the Mountain Wind!ountain Wind II pair of wind sites.) After determining the set of NRL sites to be used in the simulation analysis, NREL data were formatted, compiled by site, and labeled using their PacifiCorp counterpart's name. Similar to the earlier approach in formatting the PacifiCorp data, NREL wind output data were converted into capacity factor terms (using a 30 MW capacity value for each site as specified in the NRL description of the dataset)?4 A.4.5 Pairing of Wind Profiles Used for Regression Recognizing the monthly seasonality of wind data, each modeled pair required twelve separate regression models per year, one for each month.25 To ensure the use of observed historical wind data is meaningful, we require that a full year of overlap between a fully available wind profile and a partially missing wind profie. This means that if the partially missing wind profie only had 11 months of historical data, it. was treated as a completely missing dataset and used the NREL data to help simulate the data from the period without historical data. To simplify the rest of this explanation, the fully available wind profie was a predictor and a site with partially missing or completely missing wind profile was a predicted site (because the process effectively used the available profile to "predict" the missing profie). The Study focused on two methods in estimating monthly regressions. First, for sites with partially missing historical wind data that have at least 12 months of historical data, the data from afully available site was employed as the predictor (such as Foote Creek, Combine Hils, or Leaning Juniper) to estimate monthly coefficients. From the coeffcients derived in the regression estimation, the Study estimated the wind data for all the missing months. Second, for sites with partially missing data (and with less than 12 months historical data available) and sites with completely missing data, the NRL closest neighbor set of wind profies was employed. The process estimated monthly regression models between the closest NREL site to the predictor and the closest NREL site to the predicted. Then the coefficients estimated in those regressions were applied to the PacifiCorp fully available predictor data to simulate 1O-minuteoutput data for the predicted. This second approach implicitly assumed that the monthly relationships between the predictor and the predicted derived from the 2004-2006 period (using available NREL data) were applicable to the Initial Term as represented by the PacifiCorp data. Below in Figue 10A, a flow char depicts the steps described above. Table 3A depicts the pairs of wind sites with left colum containing the predictor and the right colum containing the predicted. 24 htt://www.nre1.gov/windlintegrationdatasets/about.html (accessed July 1,2010) 25 For example, if overlapping data for the predictor and the predicted are available for all of 2008 and 2009, we estimate a regression for Janua using data for that month from both 2008 and 2009. Then, the estimated coefficients from the regression wil be used to predict the output for January of 2007 using the predictor 2007 data for that month. September 1, 2010 Page 46 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 48 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall AM_RICAN eNlGYllllNGS COMPA Figure 10A. Wind generation data development flow chart. Metbods of Wind Data Simulation Method A: Sites with Partally Missing nata (with at least 12 months historica data)MethodS: Sites wih Completely Missing Data (or les th 12 months biorkal data) 2007 2008 Table 3A. Pairs of wind projects used in data simulation. Predicted Data UsedPredictor High Plains McFadden Rock River Rollng Hils Dunap Three Buttes Top of the World Goodnoe Marengo Mountain Wind Seven Mile Hil Spanish Fork Glenrock Foote Creek Foote Creek Foote Creek Foote Creek Foote Creek Foote Creek Foote Creek Leaning Juniper Combine Hils Foote Creek Foote Creek Foote Creek Foote Creek NRLlPacifiCorp NRLlPacifiCorp NRL/PacifiCorp NRL/PacifiCorp NRL/PacifiCorp NRL/PacifiCorp NRLlPacifiCorp PacifiCorp PacifiCorp PacifiCorp PacifiCorp PacifiCorp PacifiCorp September 1, 2010 Page 47 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 49 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MlOllICAN ENEllll MOl.li COMPAN A.4.6 Regression Analysis The estimation process of the Tobit regressions was identical across all sites-the six-lag model is applied to a predictor-predicted pair. After estimation, the resulting coefficients were used to generate data for the predicted profie for all missing time periods using the values of the predictor in those time periods.26 A sample of resulting regression coeffcients for one month for one pair of wind sites is shown in Table 4A below. Table 4A. Predictive capacity factor coeffcients for the simulation of Goodnoe Hils wind generation using Leaning Juniper actual generation data. Explanatory Varables Estimated Coeffcients Capacity Factor Leaning Juniper 0.841 *** (0.0744) -0.321 ** (0.130) 0.0314 (0.135) 0.0631 (0.135) 0.0597 (0.135) 0.00342 (0.130) 0.267*** (0.0744) Capacity Factor Leaning Juniper (t-l) Capacity Factor Leaning Juniper (t-2) Capacity Factor Leaning Juniper (t-3) Capacity Factor Leaning Juniper (t-4) Capacity Factor Leaning Junper (t-5) Capacity Factor Leaning Juniper (t-6) Observations 4,464 Note: Stadad errors in parentheses. *** p.:O.Ol, ** p':0.05, * p':O.1 A. 4. 7 Estimate Mean Values of the Predicted In general, using the estimated regression coefficients to derive a prediction for the dependent variable is done by using the mean values of the explanatory varables to arrve at the predicted mean value of the dependent variable. In this case, however, we are interested in generating predicted values of the dependent varable (predicted) for all individually observed values of the independent varable (predictor). As a result, applying the estimated regression coefficients to each individual observation of the explanatory variables will result in predicted values of the 26 Again, all estimation procedures and simulations were conducted using the commercially-available statistical softare package STATA~ (htt://www.stata.com) September 1, 2010 Page 48 of 63 PACIF.ICORP Rocky Mountain Power Exhibit No. 36 Page 50 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MlDAMERlN ¡¡IIElGY OOUlNGS COPAN predicted that are significantly less variable than the tre unobserved predicted series. This is due to the fact that the regression model assumes that the regression error is zero on average across the observations, but not in every individual instance. An ilustrative comparson of the predicted mean value to the historical actual of the same period is shown in Figue llA. Figure l1A. Comparison of actual Goodnoe Hils capacity factors with predicted mean Goodnoe Hils capacity factors derived off of Leaning Juniper generation data. Actualarf'ediìeiiGapacit Factiin 90% 80% 70%.. ~ r: t'.0I:Q.I:U 60%. 50% 40%. 30% 20% 10% 0%M.. ~'.. ~e~.COt!to ee.:0\ee ~~"l oo~.. COee ~~"l ~CO COee .~,~"l ..M c-oo~""e ....=-a .;....0\COCOCOecieee-=e e ~~ ~.~,...~.~.~.~"l "l"l "l A.4.8 Calculating the Regression Residuals To address the loss of variabilty by simply using the regression coefficients in the estimation, the technical advisor subtracted the predicted values of the dependent variable from their correspondig observed values over the overlapping subset of predicted/predictor data used for the regression estimation.27 This produced a set of regression residuals, which represent the amount by which predicted values for the known (historical) part of the data set were different from the actual observed values of the predicted. Then, each regression residual value was categorized according to the level of predicted output it was originally associated with. The predicted values are then grouped in bins of 10 percentage points to create 10 bins that cover the range of 0% to 100% capacity factor output. For example, 27 In the case of the PacifiCorp sourced data, this is done over the monthly regression data. For the Hybrid approach where NRL data was required, ths is done with the NRL data. September 1, 2010 Page 49 of 63 Rocky Mountain Power Exhibit No. 36 Page 51 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall all residuals that were associated with a predicted output between 10% and 20% are grouped together. As Figues 12A and 13A show, the distrbutions of those residuals var across bins. Figure 12A. Highly non-normal residuals from bin 5 of the March regression of Goodnoe Hils ca aci factor derived from observed LeaDing Juniper data. Figure 13A. Highly non-normal residuals from bin 7 of the March regression of Goodnoe Hils ca acity factor derived from observed Leanin Juniper data. September 1, 2010 Page 50 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 52 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MlO£RlN ENElGY OOi.GS COPANY A. 4. 9 Sample of Residuals According to Simulated Output Ranges The next step involved randomly drawing residuals from the previously defmed bins and "adding them back" to the simulated mean 10-minute wind output. The procedure of makin§ random draws from an empirical distrbution of residuals is called "bootstrapping" residuals. 8 In the context of this study, the technical advisor applied the bootstrapping procedure by randomly drawing29 a residual from a corresponding bin and adding it to the predicted mean capacity factor value. For example, if a predicted capacity factor value for a missing data point falls within the 10% to 20% interval, a residual value wil be randomly drawn from the bin that contains the residuals of the corresponding capacity factor of the historical data when compared with the simulated (or predicted) mean values. A.4.10 Application of a Non-Linear 3-Step Median Smoother to the Sampled Residuals After generating a time-series of bootstrapped residuals, the additional step of applying a non- linear smoother to the series, called the "span-3 median smoother" was taken. The span-3 median smoother is a process by which the median of the curent, previous, and next period value - in this case, it is calculated by takig the median of residual(t-l), residual(t), residual( t+ 1 )30 - and using that median as the residual for the curent period. The purose of this approach is two-fold. Firstly, the median smoother ensures that the time-series of residuals resembles the time behavior of wind more closely, with lags affecting the instantaneous results. Secondly, the span-3 median smoother introduces a time-dependency to the data set, which is known to exist in the original wind data.3! The technical advisor then added the smoothed time-series of the randomly drawn residuals to the predicted mean capacity factor values for each ten-minute point; then checking the resulting data to make sure the estimates remained within the 0 - 100% capacity factor range. 28 This name alludes to the fact that, absent prior knowledge of the distribution, the researcher has to pull herself by the bootstrps by drawing randomly from the empircally-derived residual data in order to generate residuals. 29 Radom draws are done with replacement as implemented by the STAT A(t bsample procedure. 30 For example, see "STATA Base Reference Manual Release 11", Stata Corp. p. 1758; Mosteller, F. and Tukey, John W., "Data Analysis and Regression: A Second Course in Statistics", Addison-Wesley: 1977., pp. 52-58. 31 Although the non-linear smoothing approach does not exactly replicate the auto-regressive behavior of the wind data, it introduces some similar dependency~ September 1, 2010 Page 51 of 63 PAC. I FICORP Rocky Mountain Power Exhibit No. 36 Page 53 of 63 Case No. PAC.E-11-12 Witness: Gregory N. Duvall A Mlti.RlCN e_VlINll COMPAV AppendixB Regression Coefficients and Relative Significance Regression Results by Month for Glenrock Predicted by Foo Creek Capacity Factor Foote Crek (t) Capacity Factor Foote Crek (t-l) Capacity Factor Foote Crek (t-2) Capacity Factor Foote Creek (t-3) Capacity Factor Foote Crek (t-4) Capacity Factor Foote Creek (t-5) Capacity Factor Foote Creek(t-6 Number of Observations Note: Standard errs in parentheses. *.. p""O.OI, .. p""O.05, · p""O.1 Regression Results by Month for Spash Fork Predcte by Foote Creek lanatory Varibles Capacity Factor Foote Crek (t) Capacity Factor Foote Creek (t-I) Capacity Factor Foote Crek (t-2) Capacity Factor Foote Creek (t-3) Capacity Factor Foote Creek (t-4) Capacity Factor Foote Creek (t-5) Capacity Factor Foote Creek (t-6) Number of Observations Note: Standad errrs in paretheses. *.. p""O.OI,.. p"".05,. p"".1 September 1, 2010 Page 52 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 54 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MlOERlAN .NEGY IlGS COMPA Regression Results by Month for Sewn Mle mil Preilcted by Foo Creek Capacity Factor Foote Crk (t) Capacity Factor Foote Oeek(t-l) Capacity Factor Foote Crk (t-2) Capacity Factor Foote Oeek (t-3) Capacity Factor Foote Oeek (t-4) Capacity Factor Foote Creek (t-5) Capacity Factor Foote Oeek (t-6) Number of Obserations Note: Standar enors in partheses. ... p"Ü.Ol, .. p"Ü.05, · p"Ü.l Regression Results by Month for Mountan Wind Predicte by Foote Creek Capacity Factor Foote Creek (t) Capacity Factor Foote Oeek (t.1) Capacity Factor Foote Creek (t-2) Capacity Factor Foote Oeek (t-3) Capacity Factor Foote Creek (t-4) Capacity Factor Foote Creek (t-5) Capacity Factor Foote Creek (t-6 Number of Observations Note: Standar enors in parentheses. ... p"Ü.Ol, .. p"Ü.05, · p"Ü.l September 1, 2010 Page 53 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 55 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A M~ICN EM111GY lIOlNGS CO Regression Resnlts by Month for Maengo Predicted by Conine Hills Capacit Factor Combine Hi (t) Capacity Factor Combine Hi (t-l) Capacity Factor Combine Hi (t-2J Capacit Factor Conbine Hi (t-3J Capacity Factor Conbine His (t4J Capacity Factor Conbine Hi (t-5) Capacit Factor Conbine His (t-6 Number ofObseivations Note: Standard errs in paretheses. ... p"O.OI, .. p.q.05, · p"O.i Regression Resnlts by Month for Gooe Predicted by Leanng Jnniper Capacit Factor Leaning Juniper (tJ Capacit Factor Leaning Juniper (t-l J Capacit Factor Leaning Juniper (t-2J Capacity Factor Leaning Juniper (t-3J Capacit Factor Leaning Juniper (t4J Capacity Factor Leaning Juniper (t-5J Capacity Factor Leaning Juniper (t-6J Number of Obserations Note: Standar errrs in parentheses. ... p"O.OI, .. p"O.05, . pai.l September 1,2010 Page 54 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 56 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MIlMMERtC ¡NfllGY lIUl1lGS COM Regression Resulls by Month for Top of the WorldPredctedby Foo Creek Capacity Factor Foote Crek (t) Capacity Factor Foote Crek (t-l) Capacity Factor Foote Crek (t-2) Capacity Factor Foote Crek (t-3) Capacity Factor Foote Crek (t-4) Capacity Factor Foote Crek (t-5) Capacity Factor Foote Crek (t-6 Nurrer ofObselvations Note: Standard errrs in parentheses. ... p.:O.Ol, .. p.:.05, · p.:.l Regression Results by Month for Thee Buttes Predcted by Foo Creek Exlanato Varbles Capacity Factor Foote Crek (t) Capacity Factor Foote Crek (t-l) Capacity Factor Foote Crek (t-2) Capacity Factor Foote Crek (t-3) Capacit Factor Foote Crek (t-4) Capacit Factor Foote Crek (t-5) Capacit Factor Foote Crek (t-6 Number ofObselvations Note: Standard errs in parntheses. ... p.:.Ol," 1'.05,' p.:.l September 1, 2010 Page 55 of 63 Rocky Mountain Power Exhibit No. 36 Page 57 of 63 Case No. PAC-E-11-12 Witness: Gregory N. DuvallCIFICORP Regression Res u1ts by Month for Dunlap Predicted by Foote Creek Capacit Factor Foote crek (t) Capacit Factor Foote crek (t-I) Capacity Factor Foote crek (t-2) Capacit Factor Foote crek (t-3) Capacit Factor Foote crek (t-4) Capacity Factor Foote crek (t-5) Capacit Factor Foote crek (t-6) Numer of Observations Note: Standar errs in parntheses. ... po(.oi, .. po(.05, . po('\ Regression Results by Month for Rolling mils Predcted by Foote Creek Janato Varbles Capacity Factor Foote crek (t) Capacity Factor Foote crek (t-I) Capacity Factor Foote crek (t-2) Capacity Factor Foote crek (t-3) Capacity Factor Foote crek (t-4) Capacity Factor Foote crek (t-5) Capacity Factor Foote crek (t-6) Numer of Obserations Note: Standar errs in partheses. ... po(.oi, .. po(.05, . p""O.\ September 1, 2010 Page 56 of 63 Rocky Mountain Power Exhibit No. 36 Page 58 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall Regression Res ults by Month for Rock Rher Predicted by Foote Creek Capacit Factor Foote crek (t) Capacit Factor Foote crek (t-!) Capacity Factor Foote crek (t-2) Capacit Factor Foote crek (t-3) Capacit Factor Foote crek (t-4J Capacit Factor Foote crek (t-5J Capacit Factor Foote crek (t-6) Number of Observations Note: Standard errrs in parentheses. ... po.oi, .. po(.05, . po('\ Regression Results by Month for McFaddn Predicted by Foote Creek Janato Varbles Capacity Factor Foote crek(t) Capacity Factor Foote crek (t-I) Capacity Factor Foote crek (t-2) Capacit Factor Foote crek (t-3) Capacity Factor Foote crek (t-4J Caacit Factor Foote crek (t-5J Caacity Factor Foote crek (1-6) Number of Observations Note: Stadar errrs in partheses. ... po(.oi, .. p""O.05, · po('\ September 1, 2010 Page 57 of 63 PACIFICORP Rocky Mountain Power Exhibit No. 36 Page 59 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall AM_lllC ElIVlIl1GS COMf' Regression Resnlts by Month for mgb Plains Predicted by Foote Creek Capacit Factor Foote crek (t) Caacity Factor Foote crek (i-i) Capacity Factor Foote crek (t-2) Capacity Factor Foote crek (t-3) Capacity Factor Foote crek (t-4) Capacity Factor Foote crek (t-5) Caacity Factor Foote crek (t-6 Number of Observations Note: Standar errrs in parntheses. ... p""O.OI, .. po(.05, · po('\ September 1, 2010 Page 58 of 63 Rocky Mountain Power Exhibit No. 36 Page 60 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall Appendix C Operating Reserve Demand Seasonal Detail This Appendix presents the monthly component operating reserve servce demand calculated for the PacifiCorp East and West Balancing Authority Areas in the Study. The 1,372 MW and 1,833 MW penetration scenarios include some simulated wind data; the load-only and 425 MW penetration scenaros do not. Table Cl.West Balancing Authority Area, Load Load Following Regulation~Down ~Down January 127 129 125 82 February 93 103 111 73 March 114 115 109 77 April 84 87 103 65 May 93 101 95 72 June 82 83 78 63 July 93 96 69 64 August 79 84 65 60 September 96 104 88 64 October 83 83 98 62 November 149 166 127 95 December 125 116 101 86 Only Table C2.West Balancine: Authority Area, 425 Load Following Regulation ~Down ~~ January 132 134 131 91 February 104 110 117 82 March 128 124 118 92 April 96 96 110 78 May 108 109 102 84 June 103 96 88 80 July 110 105 78 79 August 98 94 76 77 September 105 107 94 73 October 97 88 104 74 November 157 169 133 103 December 132 121 106 94 MW Table C3. West Balancing Authority area, 1,372 MW September 1, 2010 Page 59 of 63 ~CIFICORP Rocky Mountain Power Exhibit No. 36 Page 61 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall load Following Regulation ~Down ~Down January 153 150 171 139 February 122 122 152 129 March 160 152 152 140 April 133 122 150 121 May 135 131 136 123 June 131 123 127 118 July 128 122 110 104 August 118 113 103 104 September 125 121 118 101 October 124 105 126 104 November 181 180 152 131 December 159 138 142 131 Table C4. West Balancine: Authority area, 1,833 Load Following Regulation ~Down iJ Down January 153 150 171 139 February 124 124 152 129 March 162 154 152 140 April 136 123 150 121 May 137 133 136 123 June 133 125 127 118 July 129 123 110 104 August 120 115 103 104 September 126 122 118 101 October 125 106 126 104 November 182 180 152 131 December 161 139 142 131 MW September 1, 2010 Page 60 of 63 Rocky Mountain Power Exhibit No. 36 Page 62 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall Table C5. East Balancine: Authority area, Load 0 load Following Regulation ~Down ~Down January 127 131 150 110 February 117 122 131 98 March 135 138 122 102 April 105 103 145 95 May 146 145 133 114 June 143 152 134 114 July 157 155 130 112 August 162 162 122 111 September 144 162 127 105 October 139 146 116 97 November 154 164 161 110 December 145 149 182 112 nly T bl C6 E tB I A th .t A 425MWae.as a ancine:u oruy rea, load Following Regulation~Down ~Down January 132 135 152 113 February 120 125 134 101 March 139 142 124 105 April 112 107 148 99 May 151 148 137 118 June 148 155 137 118 July 161 157 132 115 August 165 164 124 114 September 149 165 130 109 October 143 150 119 101 November 158 168 163 113 December 150 154 185 116 September 1, 2010 Page 61 of 63 RP Rocky Mountain Power Exhibit No. 36 Page 63 of 63 Case No. PAC-E-11-12 Witness: Gregory N. Duvall A MltillICAI MWTable C7. East Balancing Authority Area, 1,372 load Following Regulation~Down iJ Down January 187 193 201 175 February 201 195 210 189 March 212 209 207 200 April 193 174 212 182 May 204 184 183 179 June 205 192 189 185 July 205 177 170 172 August 204 187 164 166 September 219 203 185 177 October 218 211 202 192 November 230 227 232 197 December 212 228 253 207 C MWTable8. East Balancing Authority area, 1,833 Load Following Regulation~Down iJ pown January 240 262 250 241 February 256 262 264 247 March 247 247 235 236 April 236 213 243 223 May 228 205 203 202 June 232 210 204 202 July 220 185 177 183 August 216 197 176 179 September 245 222 201 199 October 257 251 235 230 November 276 290 279 259 December 291 299 300 266 September 1, 2010 Page 62 of 63