20100528Eelkema Direct.pdf

20\ß MAY 28 PM 12: 06
BEFORE THE IDAHO PUBLIC UTILITIES COMMISSION
IN THE MATTER OF THE )
APPLICATION OF ROCKY )
MOUNTAIN POWER FOR )
APPROVAL OF CHANGES TO ITS )
ELECTRIC SERVICE SCHEDULES )
AND A PRICE INCREASE OF $27.7 )
MILLION, OR APPROXIMATELY )13.7 PERCENT )
CASE NO. PAC-E-10-07
Direct Testimony of Peter C. Eelkema
ROCKY MOUNTAIN POWER
CASE NO. PAC-E-10-07
May 2010
1 Q.
2
3 A.
Please state your name, business address and present position with Rocky
Mountain Power ("Company").
My name is Peter C. Eelkema, my business address is 825 N.E. Multnomah, Suite
4 600, Portland, Oregon 97232, and my present position is Lead/Senior Consultant,
5 Load and Revenue Forecasting.
6 Qualifications
7 Q.
8 A.
9
10
11
12
13
14
15
16
17
18
19
20 Q.
21 A.
22
23
Briefly describe your educational and professional background.
I received an undergraduate degree in Economics from San Jose State University
in San Jose, California. I also received a PhD in Economics from the University
of Kansas.
From September 1989 to October 1993, I was a Managing Research
Economist at the Kansas Corporation Commssion. From October 1993 to March
1996, I was an Economist at the Nevada Office of Advocate for Customers of
Public Utilities. From March 1996 to March 1998, I was a Senior Economist,
Forecasting, at Sierra Pacific Power/Nevada Power Company, and from March
1998 to January 2005, I was a Staff Economist, Forecasting at Sierra Pacific
Power/Nevada Power Company. From January 2005 to May 2008, I was a
Consultant, Load and Revenue Forecasting at PacifiCorp. I was promoted to my
. current position in May 2008.
Please describe your current duties.
I am the senior consultant of the Load and Revenue Forecasting group. The Load
and Revenue Forecasting group is responsible for the development of the test year
kilowatt-hour ("kWh") sales, number of customers, system loads, and system
Eelkema, Di - 1
Rocky Mountain Power
1
2 Q.
3 A.
peaks for the Company's six retail jurisdictions.
Have you previously testifed before a regulatory commission?
Yes. I have testified before the Utah, Wyoming, and Nevada Public Service
4 Commssions, and the Kansas Corporation Commssion.
5 Purpose and Summary of Testimony
6 Q.
7 A.
8
9
10
11
12
13
14
15 Q.
16
17 A.
18
19
20
21
22
23
Please explain the purpose of your testimony in this proceeding.
I describe how PacifiCorp developed the test year number of customers and bils,
kWh sales at the meter ("sales"), and system loads and system peak loads at the
system input level ("loads") for the 12-months ending December 31, 2010. The
Company produces test year sales and peak for all six states in which the
Company serves retail customers which are necessar for the development of
jurisdictional allocation factors, test year revenues, and test year net power costs.
In addition to the test year bils and sales at the class level, the Company has
developed test year bils and kWh sales by rate schedule for Idaho.
How were 2010 test year sales utilzed in preparation of this general rate
case?
Test year loads for Idaho for the twelve-months endig December 31,2010, were
used to calculate net power costs, and also used by Company witness Mr. Steven
R. McDougal to calculate the revenue requirement and jursdictional allocation
factors. Additionally, test year sales by rate schedule are used by Company
witnesses Mr. Wiliam R. Griffith and Mr. C. Craig Paice to allocate costs between
customer classes and to design rates which correctly reflect the cost of service.
The sum of energy by rate schedule ties to test year energy by customer class.
Eelkema, Di - 2
Rocky Mountain Power
1 Q.
2
3 A.
4
5
6
7
8
9
10
Please explain why this Commission should rely on a 2010 test year sales
instead of 2009 actual sales and peak as the basis for the rate case?
Idao's 2009 sales were unusual for at least two reasons. First, 2009 industrial
sales were abnormally low. Second, an unusually wet sprig resulted in a
decrease in irgation sales. The usage from these two customer classes decreased
approximately 20 to 21 percent from 2007 and 2008 level, reducing Idaho's 2009
total sales approximately 12 to 13 percent from 2007 or 2008 levels. Table 1
provides a comparson of weather normlized sales for the past thee years and
the test year for the Idaho jurisdction.
Table i - Idaho Historical Weather Normalized and Test Year Sales (MWh)
Idaho
2007 2008 2009 2010
.Actual Actual Actual Test Year
Residential 703,206 707,545 701,865 708,442
Commercial 395,197 394,786 432,685 402,252
Industrial 1,667,149 1,642,706 1,301,528 1,645,256
Irgation 636,390 619,643 519,126 545,290
Lighting 2,215 2,488 2,556 2,560
Total 3,404,156 3,367,168 2,957,760 3,303,800
11 Table 1 shows Idaho test period sales are 11.7 percent higher than weather
12 normlized sales in 2009, but in line with 2007 and 2008 data. Below, Table 2
13 provides the same view at a total Company leveL.
Eelkema, Di - 3
Rocky Mountain Power
1 Table 2 - Company Historical Weather Normalized and Test Year Sales (MWh)
Total Company
2007 2008 2009 2010
Actual Actual Actual Test Year
Residential 15,553,797 15,746,851 15,614,657 15,852,572
Commercial 15,788,129 15,956,997 16,084,321 16,104,632
Industrial 19,396,932 20,124,653 18,711,760 18,959,206
Irgation 1,398,259 1,366,554 1,240,038 1,285,620
Public Authority 428,141 447,396 437,218 436,640
Lighting 136,080 141,122 144,765 141,150
Total 52,701,339 53,783,573 52,232,759 52,779,820
2 Table 2 shows tota Company sales for the test period are 1.0 percent higher than
3 weather normalized sales in 2009.
4 Q.How is your testimony organized?
5 A.My testimony explains how I developed the normalized load used in this case. I
6 wil address thee main areas: (1) I describe the process for developing test year
7 sales for residential, commercial, irgation, lighting, and industral customer
8 classes; (2) I describe the process of adding line losses and then spreading the
9 load to each hour of the test year; and (3) I describe the process of developing test
10 year sales and bils by rate schedule.
11 Summary of Development of Test Year Loads
12 Q.Please provide a general overview of the methodology to develop tet year
13 sales and peak.
14 A.In summ, this methodology consists of using monthly actual sales by customer
15 class in a model which fits sales to weather and economic drvers. Then using
16 economic drvers, the model yields monthly test year sales by customer class. In
Eelkema, Di - 4
Rocky Mountain Power
1 a separate analysis, the model uses sales and peak data along with weather
2 varables to yield jursdictional test year peaks. Test year monthly sales become
3 the basis of test year loads by adding line losses, i.e., kWh sales levels are
4 grossed-up to a generation or "input" leveL. The monthly loads are then spread
5 out to each hour based on the test period peaks and actual hourly load.
6 Test Year Sales for Non-Industrial Customer Classes
7 Q.
8 A.
9
10
11 Q.
12 A.
13
14
15
16
17
18
19
20
21 Q.
22 A.
23
How are test year monthly sales developed by customer class?
The Company developed test year monthly sales as a product of test year number
of customers and test year sales per customer. The Company uses this
methodology for all customer classes except for the industrial customer class.
How is the test year number of customers developed?
The Company develops the test year number of customers using regression
models based on the Januar 1997 to Januar 2010 time period. The Company
also used the most recently available economic drvers from IHS Global Insights,
which were released in December 2009. For the residential class, test year
number of customers uses IHS Global Insight's test year number of households as
the major driver. For the commercial class, test year number of customers uses the
test year number of residential customers as the major drver. For irgation and
street lighting.classes, the number of customers is faily static and developed
using regression models without any economic drvers.
How is test year average use per customer for customer classes developed?
The Company models sales per customer for the residential class through a
Statistically Adjusted End-use ("SAE") model, which combines the end-use
Eelkema, Di - 5
Rocky Mountain Power
1 modeling concepts with traditional regression analysis techniques. Major drvers
2 of the SAE-based residential model are heating and cooling related varables,
3 equipment shares, satuation levels and efficiency trends, and economic drivers
4 such as household size, income and energy price.
5 For the commercial class, the Company develops test year sales per
6 customer using regression analysis techniques with employment used as the major
7 economic drver in addition to weather-related varables.
8 For other classes, the Company develops test year sales per customer
9 through regression analysis techniques using time trend varables.
10 Test Year Industrial Class Sales
11 Q.
12
13 A.
14
15
16
17
18
19
20
21
22
23
How does the Company develop test year sales for the industrial customer
class?
The industral customers are separated into three categories: (1) existing
customers that are monitored by the Customer Community Managers ("CCMs");
(2) new large customers or expansions by existing large customers; and (3)
industral customers that are not assigned CCMs. Customers are assigned CCMs
if they have a peak load of one megawatt or more at a single site.
The Company develops test year sales for the first two categories though
the data gathered by the CCMs assigned to each customer. The CCMs have
ongoing diect contact with large customers and are il the best position to know
about the customer's plans for changes in business processes, which might impact
their energy consumption.
The Company develops the porton of test year industral sales related to
Eelkema, Di - 6
Rocky Mountain Power
1
2
3
4
5
6
7
8 Q.
9
10 A.
11
12
13
14
15
16
17
18
new large customers and expansion by existing large customers based on direct
input of the customers, load factors, and the probabilty of the project occurrence.
Smaller industrial customers, i.e., under one megawatt, ar more homogeneous
. and are modeled using regression analysis with trend and economic varables.
Employment is used as the major economic drver.
The Company develops the total test year industrial sales by aggregating
test year sales for the thee industral customer categories.
Why does the Company develop test year industrial sales using a different
methodology than the methodology used for the other customer classes?
The Company relies on a different methodology because of the diverse makeup of
the customers within the class. In the industral class, there is no "typical"
customer. Large customers have very diverse usage patterns and power
requirements. It is not unusual for the entire class to be strongly influenced by the
behavior of one customer or a small group of customers.
The non-industral customer classes are generally composed of many
smaller customers that have simiar behaviors and usage patterns. No small group
of customers, or single customer, influences the movement of the entie class.
This difference requires the different processes for developing test year sales.
19 Development of Test Year Hourly Loads
20 Q.
21 A.
22
23
Please outline the development of test year hourly loads.
After the Company develops test year monthly energy sales by customer class,
test year hourly loads is developed in two steps:
First, monthly and seasonal peaks for each state are developed. The
Eelkema, Di - 7
Rocky Mountain Power
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15 Q.
16 A.
17
18
monthly. peak model uses historic peak-producing. weather for each state, and
incorporates the impact of weather on peak loads though several weather
varables which drve heating and cooling usage. These weather varables include
the average temperatue on the peak day and lagged average temperatures. Test
year peaks based on average monthly historical peak-producing weather for the
period 1990-2009.
Second, the Company obtains test year hourly loads for each state from
hourly load models using state-specific hourly load data and daily weather
varables. The Company develops hourly loads using a model that incorporates
the twenty-year average temperatues, a typical annual weather pattern, and day-
type varables such as weekends and holidays. The hourly loads are calibrated to
match the monthly and seasonal peaks from the first step above. Also, the hourly
loads are calibrated so that the monthly sum of hourly loads equals monthly sales
plus line losses.
How are monthly system coincident peaks derived?
After the test year hourly loads are developed for each state, hourly loads are
aggregated to the total system leveL. The system coincident peaks can then be
identified as well as the contribution of each jurisdiction to those monthly peaks.
19 Curtailment Adjustments
20 Q.
21
22 A.
23
Please describe test year curtailment and how it is reflected in the model
driven results?
Test year curailment is developed for industral sales and peak in Idaho and Utah.
Industral curailment consists of interrption and buy-th. Test year curtilment
Eelkema, Di - 8
Rocky Mountain Power
1
2
3
4
5
6
7
8
9
10
11 Q.
12
13 A.
14
15
16 Q.
17 A.
18
19
20
21
22
23
is developed for irgation peak in Idaho and Utah. Irgation curtment consists
of only interrption. Test year residential air conditioning curailment is
developed for Utah.
Test year sales are gross of buy-th, but net of interrption. That is, the
model drven test year sales are drven by historic sales data which contains both
retail sales and buy-th.
The historical monthly peak data is also gross of buy-thr, but net of
interrption. Monthly historical peaks have been adjusted to reflect the
interrption which occurred at that hour; therefore, the model drven peaks is
gross of interrption and buy-thr.
Why didn't you make one adjustment to reflect curtailment in the model
driven results?
As mentioned earlier, test year sales and peak are used by different groups and
each group treats curtailment differently. Therefore, it is necessary for each group
to make an adjustment for curtailment specific to how they use curailment.
Please describe the method used to develop test year curilment.
Test year sales is not adjusted to reflect the effects of the irgation program
because the Company is assuming 100 percent take-back. Irgators knowing they
may be curailed durng the 2:00 p.m. to 6:00 p.m. window wil irgate around
the curtilment hours. As a result, energy wil be shifted away from curailment
hours, but the daily energy wil not change appreciably. Also, the Utah residential
ai conditioner program is energy neutral.
Test year peak curailment to reflect the effects of the irgation programs
Eelkema, Di - 9
Rocky Mountai Power
1 in both Idaho and Uta are calculated as the ratio of class usage at the time of
2 coincident peak to maximum class usage times the maimum curailment from
3 tbe program.
4 Test year peak and sales curailment to reflect the effects of the Idaho and
5 Utah industrial programs is estimated based on a 5-year average historical
6 monthly interrption and buy-th. In 2010, there is an increase in Idaho
7 industral potential curtailment which is not reflected in the historical data. It is
8 assumed that the customer wil buy-thr these additional hours.
9 Test Year Rate Schedule Sales
10 Q.
11 A.
12
13
14
15
16
17 Q.
18 A.
19
20
21
22 Q.
23 A.
Why does the Company develop test year rate schedule sales?
To perform the class cost of service and rate design analysis, two additional
projections that are based on test year kWh sales and test year number of
customers are required. Once the test year kWh sales are complete, it must be
applied to individual rate schedules to develop test year kWh sales by rate
schedule. In addition, test year number of customers must be expressed in
number of bils.
How are test year rate schedule level sales projected?
This process is cared out in two steps. First, the Company projects each rate
schedule's share of the customer class sales. Second, the Company multiplies the
projected rate schedule share by the test year customer class sales to project sales
by rate schedule.
How is the number of test year bils for each schedule developed?
Simiar to the test year rate schedule sales, the test year rate schedule bils is
Eelkema, Di - 10
Rocky Mounta Power
1 cared out in several steps. First, the Company calculates the ratio of bils to
2 sales by rate schedule to bils by customer class. Second, this ratio is projected
3 for the test period based on the regression results. Third, the ratio is multiplied by
4 the customer class bils to produce the bils by rate schedule.
5 Conclusion
6 Q.Do you consider the test year sales and loads to be reasonable?
7 A.Yes. Actual 2010 sales have an equal probabilty of being more than or less than
8 test year sales.
9 Q.Does this conclude your testimony?
10 A.Yes.
Eelkema, Di - 11
Rocky Mountai Power