20140617June 2014 Solar Integration Study.pdf

Rr'c'[lvi- 1')
?$1\ JUH l', Pt{ 3: 0B
ul J,?H L*cir,q,.1 iE s r i; i',
3rffi*.
AN IDACORP COMDANV
June 17,2014
VIA HAND DELIVERY
Jean D. Jewell, Secretary
ldaho Public Utilities Commission
472 West Washington Street
Boise, ldaho 83720
Re: June 2014 Solar lntegration Study Report
Dear Ms. Jewell:
ln Order No. 33043, Case No. IPC-E-14-09, the ldaho Public Utilities
Commission ("Commission") directed ldaho Power Company ("ldaho Powed' or
"Company") to "complete its solar integration study as soon as possible." The study is
complete. Enclosed please find five (5) copies of ldaho Powe/s June 2014 Solar
lntegration Study Report.
ldaho Power is preparing an application case filing where it will ask the
Commission to implement a solar integration charge based upon this study. The
Company anticipates this filing will be made within the next two weeks. Please contact
me at [phone] if you have any questions.
DONOVAN E. WALKER
Lead Counsel
DEW:csb
Enclosures
,5L?c/a
Donovan E. Walker
1221 W. ldaho St. (83702)
PO. Box 70
Boise, lD 83707
i'
r E C.!: l\i -.i)I '.:- v ''* -
?01\ JUi{ l1 PH 31 0B
Solar lntegration Study Report
rem.
June 2014
@ 2014 Idaho Power
ul lifii[$-co-i*:ti!s
ldaho Power Company Solar lntegration Study Report
Page 1
Solar lntegration Study Report ldaho Power Company
LIST OF TABLES
Table I
Solar build-out scenarios studied ......................3
Table 2
Average integration cost per MWh for solar build-out scenarios.... ..........................3
Table 3
Solar build-out scenarios studied ......................6
Table 4
AgriMet site latitude, longitude, and elevation used in IPC's solar integration study .................7
Table 5
Forecast error for the hour-ahead solar production forecast...... ...........10
Table 6
Forecasted incremental and decremental capacity held in reserve, water year 2012 .................1 I
Table 7
Inputs for the solar integration study production cost simulations............ ..............12
Table 8
Average integration cost per MWh for solar build-out scenarios.... ........................15
Table 9
Incremental integration cost results for solar build-out scenarios ........15
LIST OF FIGURES
Figure I
AgriMet sites used in IPC's solar integration study ............7
LIsT OF APPENDICES
Appendix I
Solar integration study appendix.... .................21
Page2
ldaho Power Gompany Solar lntegration Study Report
ExecuTME SUMMARY
Electric power from solar photovoltaic resources exhibits greater variability and uncertainty
than energy from conventional generators. The greater variability and uncertainty exhibited by
solar photovoltaic resources require an electric utility integrating solar to modiS the operation of
dispatchable generating resources. The modified operation involves the sub-optimal dispatch of
generators to carry extra capacity in resenre for responding to unplanned solar excursions.
The objective of the Idaho Power solar integration study is to determine the costs of the
operational modifications necessary to integrate solar photovoltaic plant generation. This study
determines these costs for four solar build-out scenarios provided in Table l.
Table 1
Solar build-out scenarios studied
!nstalled Capacity of Solar Build-Out Scenarios
Site 100 megawatts (MW)300 Mw 500 Mw 700 Mw
Parma, lD
Boise, lD
Grand View, lD
Twin Falls, lD
Picabo, lD
Aberdeen, lD
TotalMW
10
20
20
20
10
20
r00
30
60
60
60
30
60
300
50
100
100
100
50
100
500
100
100
150
100
100
150
700
The study determines solar integration costs through paired simulations of the ldaho Power
system for each solar build-out scenario. Each pair of simulations consists of a test case in which
extra capacity in reserve is required ofdispatchable generators to allow them to respond to
unplanned solar excursions and a base case in which no extra capacity in reserve is required.
The solar integration costs indicated by the simulations are provided in Table 2.
Table 2
Average integration cost per MWh for solar build-out scenarios
0-100 Mw 0-300 Mw 0-s00 Mw 0-700 Mw
!ntegration cost $0.4O/]'/VVh $1.2olMwh $1.80/[/VVh $2.5O/tvlvvh
Note: Costs arein 2014 dollars and rounded from simulation results to the nearest $0.'10.
Page 3
Solar lntegration Study Report ldaho Power Company
This page left blank intentionally.
Page 4
ldaho Power Company Solar lntegration Study Report
AcxT.IoWLEDGMENTS
Idaho Power acknowledges the important contribution of the Technical Review Committee
(TRC) in this solar integration study. The TRC has been involved from the study outset in
August 2013 and has provided substantial guidance. [daho Power especially thanks the TRC for
the collegial discussions of solar integration during TRC meetings. These discussions helped
shape the study methods followed and are consistent with the TRC guidelines as provided by the
Utility Variable-Generation lntegration Group (WIG) and the National Renewable Energy
Laboratory NREL) GIVIG and NREL n.d.). The following are members of the Idaho Power
solar integration study TRC:
o Brian Johnson, University of Idaho
o Jimmy Lindsay, Portland General Electric (formerly of Renewable Northwest Project)
o Kurt Myers, Idaho National Laboratory
o Paul Woods, (formerly of City of Boise)
o Cameron Yourkowski, Renewable Northwest Project (replacing Jimmy Lindsay)
Staff from the ldaho and Oregon regulatory commissions have participated as observers
throughout the process. The following staff have been observers of the process:
o Brittany Andrus, Public Utility Commission of Oregon (OPUC) staff
o John Crider, OPUC staff
o Rick Sterling,Idaho Public Utilities Commission (IPUC) staff
TRC members and regulatory observers serve either voluntarily or are paid by their own
employers and receive no compensation from ldaho Power. The company is grateful for the
TRC's time spent supporting the study and recognizes this support has led to a better study.
lrurnooucnoN
Electric power from solar photovoltaic resources exhibits greater variability and uncertainty
than energy from conventional generators. Because of the greater variability and uncertainty,
electric utilities incur increased costs when their other generators are called on to integrate
photovoltaic solar plant generation. These costs occur because power systems are operated less
optimally in order to successfully integrate solar plant generation without compromising the
reliable delivery of electrical power to customers. Idaho Power has studied the modifications it
must make to power system operations to integrate solar photovoltaic power plant generation
connecting to its system. The objective of this solar integration study is to determine the costs of
the operational modifications necessary to integrate solar plant generation. This report is
intended to describe the operational modifications and the resulting costs.
Page 5
Solar lntegration Study Report ldaho Power Company
In collaboration with the TRC, Idaho Power organized the study into four primary steps:
1. Data gathering and scenario development
2. Statistical-based analysis of solar characteristics
3. Production cost simulation analysis
4. Study conclusions and results
These steps were formulated based on an article published by the Institute of Electrical and
Electronics Engineers (IEEE) describing methods for studying wind integration (Ela et al. 2009).
While the IEEE article, which was authored by leading researchers at the NREL, was written
from the perspective of studying grid integration of wind generation, the principles underlying
the study of wind integration are readily transferrable to the study of solar integration. Both wind
and solar bring increased variability and uncertainty to power system operation, and a key
objective of an integration study for each is to understand how variability and uncertainty lead to
impacts and costs.
Dara GarnenrNc AND Scerunnro DEVELopMENT
A critical element of the solar integration study is the solar generation data developed for the
studied solar build-out scenarios. For Idaho Power's solar integration study, the solar build-out
scenarios in Table 3 were studied.
Table 3
Solar build-out scenarios studied
lnstalled Capacity of Solar Build-Out Scenarios
100 megawatts (MW)300 Mw 5OO MW 7OO MW
Parma, lD
Boise, lD
Grand View, lD
Twin Falls, lD
Picabo, lD
Aberdeen, lD
Total MW
10
20
20
20
't0
20
100
30
60
60
60
30
60
300
50
100
100
100
50
100
500
100
100
150
100
100
150
700
The above build-out scenarios were developed in consultation with the TRC to represent
geographically dispersed build-outs of solar power plant capacity. The importance of geographic
dispersion in reducing integration impacts and costs is discussed in greater detail later in this
report. The sites from the solar build-out scenarios are part of the established United States
(U.S.) Bureau of Reclamation (USBR) AgriMet Network (AgdMet). AgriMet is a satellite-based
network of automated agricultural weather stations operated and maintained by the USBR.
The stations are located in irrigated agricultural areas throughout the Pacific Northwest and are
Page 6
ldaho Power Company Solar lntegration Study Report
dedicated to regional crop water-use modeling, agricultural research, frost monitoring,
and integrated pest and fertility management. The six sites are spread across southern Idaho
and cover over 220 miles from east to west (Figure I ). Sites represent elevations ranging from
2,300 feet to 4,900 feet (Table 4).
\J*.-,t"'*fl;t.l -J'k it
ffi-
i, Parma
.i ,,
Aberdeen
l
t..
Thin Falls
Figure 1
AgriMet sites used in IPC's solar integration study
Table 4
AgriMet site latitude, longitude, and elevation used in IPC's solar integration study
Latitude (N)Longitude (west)Elevation (feet) Elevation (meter)
Parma
Boise
Grand View
Twin Falls
Picabo
Aberdeen
116.93
1 16.1 I
116.06
114.35
114.17
112.83
43.'t8
43.60
42.91
42.55
43.31
42.95
2,305
2,720
2,580
3,920
4,900
4,400
702
829
786
1 ,195
1,494
1,341
All data used in the integration study are 5-minute interval global horizontal irradiance data from
each site. Idaho Power worked directly with the USBR Pacific Northwest Region AgriMet
manager to obtain data for the sites. AgriMet data was augmented with data from the University
of Oregon Solar Radiation Monitoring Laboratory when AgriMet data was incomplete. The use
of high-resolution (5-minute interval) data is critical to characterizing the variability of solar.
An alternative data-gathering approach was necessary for the Grand View site, for which only
15-minute data was available. To acquire 5-minute data for Grand View, Idaho Power contracted
Page 7
Solar lntegration Study Report ldaho Power Company
with SolarAnywhere to provide high-resolution modeled solar data. SolarAnywhere uses hourly
satellite images processed using the most current algorithms developed and maintained by
Dr. Richard Perez at the University at Albany (SI-INrY). The algorithm extracts cloud indices
from the satellite's visible channel using a self-calibrating feedback process capable of adjusting
for arbitrary ground surfaces. The cloud indices are used to modulate physically-based radiative
transfer models describing localized clear-sky climatology.
Wavelet-Based Variability Model
AgriMet solar data represents conditions at a single point. To better reflect conditions at a solar
plant size, the TRC recommended the use of the wavelet-based variability model (WVM)
developed by Dr. Matt Lave of Sandia National Labs (Lave et al. 20l3a,b). WVM is designed for
simulating solar photovoltaic power plant output given a single irradiance point-sensor time
series. The application of the W\IM to the point-sensor time series produces a variability
reduction reflecting an upscaling of the point-source data to a solar plant-sized area.
Research and use into the WVM showed it is not useable at time steps (intervals) greater than
l0 minutes and that times steps greater than 5 minutes may under-represent variability in
dispersed systems.
Solar Plant Characteristics
This study assumes solar plants comprising the build-out scenarios occupy 7 acres per MW
of installed capacity. Solar plant sizes in the build-out scenarios, as well as figures presented for
solar generation, are in terms of AC (alternating current) MW. Photovoltaic panels are assumed
to be of standard crystalline silicon manufacture. Panels are assumed to be fixed south facing and
tilted at latitude. While panel orientation and tracking capability are key factors in the
determination of avoided costs, these attributes are of lesser importance with respect to the
variability and uncertainty driving integration costs. Illustrations and data summarizing the solar
production of the studied build-outs are provided in Appendix l.
SrnnsncAL-BASED Aruarvsrs oF Soun GnenncrERrsncs
The intent of the statistical-based analysis of solar characteristics is to translate solar's
variability and uncertainty into an increased requirement for ancillary services, where ancillary
services in this context relate to the electrical system's capacity to maintain a balance between
customer demand and generation. For the study, the variability and uncertainty associated with
solar generation were viewed from the perspective of hour-ahead scheduling of the Idaho Power
system. There are three critical elements from this perspective:
l. Forecast hourly average solar production for the operating hour being scheduled
2. Lower bound for instantaneous solar production during the operating hour
3. Upper bound for instantaneous solar production during the operating hour
From the perspective of real-time generation scheduling in practice, the lower and upper
bounds would be considered an interval or band on solar production, and the occurrence of
Page 8
ldaho Power Company Solar lntegration Study Report
solar production outside the interval at any moment during the hour is highly unlikely.
Moreover, while under prudent operating practices the occurrence of solar production outside the
lower and upper bounds should be infrequent, occasional solar excursions outside these bounds
do not necessarily bring about events for which system reliability is jeopardized. Conversely,
the occurrence of solar production within the interval between the lower and upper bounds would
be considered likely enough to warrant the scheduling of dispatchable generators to have
capacity to respond if solar production varies during the hour from the forecasted level of
production toward either bound.
An understanding of ldaho Power's participation in the regional electric power market is
critical to this approach. Idaho Power primarily participates in the Pacific Northwest's
Mid-Columbia (Mid-C) electric power market. The company participates in the Mid-C market at
multiple time frames ranging from years or months in advance for long-term operations planning
to hour-ahead generation scheduling in real time.
The focus for this study is the real-time market activities occurring as part of hour-ahead system
scheduling. The study assumes hour-ahead schedulers require the delivery of forecast hourly
average solar production and the above-described lower and upper bounds 45 minutes prior to
the start of the operating hour being scheduled. Hour-ahead scheduling is assumed binding,
and unexpected conditions occurring during the operating hour being scheduled must be
managed by changing production for Idaho Power-owned dispatchable resources.
Idaho Power recognizes efforts to establish intra-hour trading in U.S. electric power markets.
However, company experience has shown the intra-hour market to be currently highly illiquid.
Therefore, the last opportunity to participate in the electric power market is at the hour-ahead
time frame; unexpected conditions occurring during the operating hour (e.g., unexpected levels
of solar production) cannot be managed through market activity at this time.
Hour-Ahead Solar Production Forecast
The hour-ahead solar production forecast was developed to predict hourly average solar
production for the operating hour being scheduled and lower and upper bounds for
instantaneous solar production during the operating hour. This forecast was developed using a
persistence-based technique that relies on observations from the previous hours to inform the
model about subsequent forecast hours. The results ofthe forecast are a unique set of
values (average production, upper bound, and lower bound) for every hour in the year.
The average production forecast is derived based on two components. The first component
accounts for the amount of generation the system observed from the last 20 minutes of the
preceding forecast hour. This component is referred to as the persistence component.
The persistence component serves as a mechanism to increase the average forecast during times
of high solar production and decrease the average forecast during times of low solar production.
These increases and decreases are made to the forecast hourly and account for changes in solar
production. In general, the shape of the production from a solar photovoltaic system increases
before solar noon and decreases after solar noon. Every day ofthe year has a unique clear-day
shape. Generally, summer days are long and have a high potential for solar production while
winter days are shorter and have less potential. The forecast accounts for the uniqueness of each
Page 9
Solar lntegration Study Report
value for every hour in the year. The shaping component is a ratio of the maximum solar
potential of the forecast hour divided by the maximum potential of the previous hour.
By utilizing a shaping component and a persistence component, the average production
forecast captures hourly changes due to atmospheric conditions and seasonal effects. Table 5
provides the forecast error for the hour-ahead solar production forecast.
Table 5
Forecast error for the hour-ahead solar production forecast
't00 Mw 300 Mw 500 Mw 7OO MW
day by applylng an hourly shaping factor. This shaping component, or shaping factor, is a unique
Absolute Mean Hourly Error (MW)12.2
Table 5 reports the absolute mean error calculated on an hourly basis for water year 2012.
The absolute hourly error is calculated as the absolute difference between the average hourly
forecast and the average of 5-minute observed production data for a given hour. It is noted that
the 5-minute observed production data is the output of the WVM. The absolute mean hour errors
range from 1.9 MW to 12.2 MW for the 100 MW and 700 MW build-out scenarios, respectively.
The lower bound for instantaneous solar production during the operating hour is forecasted as a
percentage ofthe forecast average. tn addition to the application ofa percentage ofaverage,
the forecasting tool adjusts the lower bound forecast upward if the previous lower bound forecast
was substantially too low. As a result of this secondary adjustment to the lower bound,
the amount of incremental capacity held in reserye for the coming hour is reduced.
Similar to the lower bound, the upper bound for instantaneous solar production during the
operating hour is forecasted as a percentage ofthe forecast average. In addition to the application
of a percentage of average, the forecasting tool adjusts the upper bound forecast downward if the
previous upper bound forecast was substantially too high. As a result of this secondary
adjustrnent to the upper bound, the amount of decremental capacity held in reserve for the
coming hour is reduced.
The upper and lower bounds are expected to capture the overwhelming majority of the variability
observed in solar production. The upper bound is forecasted in such away that only 2.5 percent
of all observations exceed the upper bound for the entire year. Similarly, the lower bound is
defined in such a way that only 2.5 percent of all observations are below the lower bound for the
entire year.
The hour-ahead forecast for the average production, lower bound for instantaneous solar,
and upper bound for instantaneous solar are calculated for every hour of the year. The amount of
incremental capacity held in reserve for a given hour is calculated as the difference between the
average production forecast and the lower bound. The amount of decremental capacity held in
reserve for a given hour is calculated as the difference between the average production forecast
and the upper bound. The total amount of capacity held in reserve for a given hour is used by the
production cost model to calculate an integration cost. These reserve amounts, as well as the
hour-ahead forecast for solar production, are input to the production cost model on an hour-by-
hour basis, simulating the practice of real-time generation scheduling. Table 6 reports the
9.65.81.9
Page 10
ldaho Power Company Solar lntegration Study Report
forecasted amount of capacity held in reserve for water year 2012. Further explanation of the
derivation of the hour-ahead solar production forecast and the lower and upper bounds is
provided in Appendix l.
Table 6
Forecasted incremental and decremental capacity held in reserve, water year 2012
Solar Build.Out Scenarios
100 Mw 300 Mw 500 Mw 700 Mw
Average hourly production (MW)
Average hourly capacity held in
reservs-incremental (MW)
Average hourly capacity held in
reserve-decremental (MW)
17.0
4.9
4.9
52.5
13.2
15.2
89.0
21.2
26.9
118.2
27.6
34.8
PnooucnoN Cosr SrurulnnoN ANALysrs
The production cost simulations are designed to isolate the effects on the system associated with
integrating solar. Under this design, production cost simulations are paired into a base case and
test case, with all inputs to the paired simulations equivalent except an amount of capacity held
in reserve in the test case simulation for integrating solar. The capacity held in reserve for the test
case varies hourly depending on the hour-ahead forecast of solar production for a given operating
hour and the lower and upper bounds on instantaneous solar production for the operating hour.
The derivation of the hour-ahead solar production forecast and the lower and upper bounds is
described in the previous section of this report.
Design of Simulations
The production cost simulations are set up on a water-year calendar, where by convention a
water year is from October 1 to September 30 and is designated by the calendar year in which the
l2-month period ends. For example, water year 2013 is the l2-month period from October 1,
2012, through September 30, 2013.
The Idaho Power generating system as it exists at the time of issue of this report is assumed for
the production cost simulations. Critical elements of the simulated system of generating
resources include 17 hydroelectric facilities totaling 1,709 MW of nameplate capacity,
3 coal-fired facilities totaling 1,1l8 MW of nameplate capacity, and 3 natural gas-fired facilities
totaling 762MW of nameplate capacity. An illustration of the generating resources is provided
in Appendix 1.
Idaho Power's critical interconnections to the regional market are over the Idaho-Northwest,
Idaho-Utah (Path C), and ldaho-Montana paths. For the solar integration study modeling,
the separate paths were combined to an aggregate path for off-system access. Purchases from the
regional market are treated separately from sales to the regional market. Net firm purchases from
the market are limited on a monthly basis to only the capacity and energy required to serve
Page 11
Solar lntegration Study Report ldaho Power Company
Idaho Power's retail load. Sales to the market are limited to 500 MW in every hour. This profile
of purchases and sales reflects the current capabilities of Idaho Power's transmission system.
Idaho Power is pursuing the development of the Boardman to Hemingway Transmission Project
(B2H), which will increase ldaho Power's access to the Northwest to make additional purchases
and sales. However, the transmission line's current in-service date is at least five years into the
future. Previous integration studies have shown that unless there is a liquid capacity balancing
market, B2H will not significantly impact the solar integration cost. ldaho Power is actively
engaged in regional market discussions that could exist when B2H is completed, but the benefits
of a market are highly dependent on its design, and it is premature to speculate or incorporate in
this integration study.
Simulation lnputs
Table 7 provides key inputs to the solar integration study production cost simulations.
Table 7
lnputs for the solar integration study production cost simulations
lnput Assumed input level
Solar production
Snake River streamflows
Customer demand
Nymex-Natural gas prices
Mid-C-Electric power market prices
Non-wind PURPAI
Wind (PURPA and PPA)1
Geothermal PPAs
Waler year 2Q12
Water year 2012 (median-type streamflows)
Waleryear2012
Water year 2012
Waler year 2012
Water year 2012
Water year 2013
Waler year 2014
' PPA and PURPA represent facilities from which generation is contractually purchased as a power purchase agreement (PPA)
or under the federal Public Utility Regulatory Policies Act of 1978 (PURPA).
The selection of water year 2012 for the majority of the inputs was driven by the selection of
Snake River streamflows for water year 2012 (October 1,2}l1-September 30,2012) and the
objective to use time-synchronous input data to the greatest possible extent. Snake River Basin
streamflow conditions as observed in water year 2012 were selected because the observed water
year 2012 Brownlee reservoir inflow volume of 13.6 million acre-feet is representative of
median-type streamflow conditions. A graph of Brownlee inflow volumes for water years 1990
to 2013 is provided in Appendix 1.
The solar production data used in the production cost simulations are considered to be the solar
production that would have been observed during water year 2012had the four studied solar
build-out scenarios existed. As described previously, the solar production data is developed by
applyng a wavelet smoothing transformation technique to 5-minute interval AgriMet and
SolarAnywhere data. Importantly, the use of observed customer demand from water year 2012
allows time synchronization between solar and customer demand data in the study.
While customer demand has grown since 2012, the benefit of using time-synchronous
Page 12
ldaho Power Company Solar lntegration Study Report
customer demand and solar production data is considered to justify the use of 2012 customer
demand data. Monthly average customer demand used in the modeling is provided in
Appendix 1.
Water year 2012 Nymex natural gas prices and Mid-C electric power market prices are inputs to
the simulations. These prices, expressed as a monthly average, are provided in Appendix 1.
Wind capacity under contract with Idaho Power grew by more than 60 percent during water year
2012, expanding from 395 MW of installed capacity to 638 MW. Because of the non-constant
amount of on-line wind capacity during water year 2012,the simulations used observed hourly
wind production data for water year 2013. The amount of on-line wind capacity during water
year 2013 changed only by the addition of a single 40 MW project added during December 2013
that brought wind to the current on-line capacity of 678 MW. Monthly energy production used in
the modeling is included in Appendix l.
The remaining energy purchased from non-wind PURPA qualiffing facilities is input into the
simulations as observed during water year 2012.The monthly energy from the non-wind PURPA
facilities in included in Appendix l.
Baseload generation from geothermal facilities confractually selling to ldaho Power under PPAs
is input as curently projected from these facilities. The amount of baseload generation delivered
from these facilities varies seasonally. The amount used in the production cost simulations
ranges from22 MW to 32 MW.
Simulation Model
Idaho Power used an internally developed system operations model for the solar integration
study. The model determines optimal hourly scheduling of dispatchable hydro and thermal
generators with the objective of minimizing production costs while honoring constraints imposed
on the system. System constraints used in the model capture numerous restrictions goveming the
operation of the power system, including the following:
o Reservoir headwater constraints
o Minimum reservoir outflow constraints
o Reservoir outflow ramping rate constraints
o Generator minimum/maximum output levels
o Marketpurchase/saleconstraints
o Generator ramping rates
The model also stipulated that demand and resources were exactly in balance and importantly
that hourly reserve requirements were satisfied. The extra capacity in reserve held to manage
variability and uncertainty in solar production drives the production cost differences between the
Page 13
study's two cases. The derivation of the extra capacity in reserve held for solar is described
previously in this report.
Solar lntegration Study Report
Wind and Load Reserves
Capacity in reserve to manage variability and uncertainty in load and wind is included in the
simulations in equivalent amounts for the study's two cases. By carrying equivalent amounts in
reserve for load and wind, the production cost differences yielded by the study's simulations can
be attributed to the extra capacity held in reserve for solar. Thus, while reserves carried for load
and wind are not drivers of production cost differences in the paired simulations, it is
nevertheless desirable in simulating the system as accurately as possible to incorporate reserve
levels for load and wind representative of levels carried in practice.
To manage variability and uncertainty in load, capacity in reserve equal to 3 percent of load is
held on dispatchable generators in the modeling for the solar integration study. The amount of
simulated capacity in reserve for balancing wind is based on an analysis performed for the
Idaho Power wind integration study as described in the February 2013 Wind Integration Study
Report (Idaho Power 2013). The simulated reserves for the solar integration study are based on a
scaling of the reserves at the wind study's 800 MW wind build-out scenario to the water year
2013 wind build-out of 678 MW.
Conti ngency Reserve Obl igation
The study of integration impacts and costs focuses on the need to carry bidirectional capacity in
reserve for maintaining compliance with reliability standards. However, balancing authorities,
such as Idaho Power, are also required to carry unloaded capacity in reserve for responding to
system contingency events, which have traditionally been viewed as large and relatively
infrequent system disturbances affecting the production or transmission of power (e.g., the loss
of a major generating unit or major transmission line). System modeling for the solar integration
study imposes a contingency reserve intended to reflect this obligation equal to 3 percent of load
and 3 percent of generation, setting aside this capacity for both study cases (i.e., base and test).
Flexible Capacity Resources
As described previously, the focus of the production cost simulations for the solar integration
study is the real-time market activities occurring as part of hour-ahead system scheduling.
The study assumes hour-ahead schedulers require the delivery of forecast hourly average solar
production and the lower and upper bounds for solar production 45 minutes prior to the start
of the operating hour being scheduled. Hour-ahead scheduling is then assumed binding,
and unexpected levels of solar production occurring during the operating hour being scheduled
must be managed by Idaho Power's system.
To manage deviations in solar production from the forecast during the operating hour,
Idaho Power must schedule incremental and decremental capacity in reserve on dispatchable
generators. In the modeling for the study, this capacity in reserve is scheduled on
Hells Canyon Complex (HCC) hydroelectric generators (Brownlee, Oxbow, and Hells Canyon),
natural gas-fired generators (Langley Gulch, Danskin, and Bennett Mountain), and Jim Bridger
Page 14
ldaho Power Company Solar lntegration Study Report
coal-fired generators. The allocation of reserve to these generators matches Idaho Power's
practice for balancing variations in wind production and load.
Resurrs
The objective of the Idaho Power solar integration study is to determine the costs of the
operational modifications necessary to integrate solar photovoltaic power plant generation.
The integration costs are driven by the need to carry extra capacity in reserve to allow
bidirectional response from dispatchable generators to unplanned excursions in solar production.
The simulations performed for the ldaho Power solar integration study indicate the following
costs associated with holding the extra capacity in reserve (Table 8). The provided costs are the
costs to integrate solar production for calendar year 2014, and are not costs averaged or levelized
over the life of a solar power plant.
Table 8
Average integration cost per MWh for solar build-out scenarios
0-100 MW 0-300 Mw 0-500 MW 0-700 MW
lntegration cost $0.40/MWh $1.20l[4VVh $1.80/[A /h $2.50/MWh
Note: Costs are in 2014 dollars and rounded from simulation results to the nearest $0.10.
The integration cost results in Table 8 are the cost per MWh to integrate the full installed solar
power plant capacity at the respective scenarios studied. For example, the integtation cost results
indicate the total solar power plant capacity making up the 500 MW build-out scenario brings
about costs of $1.80 for each megawatt-hour (MWh) integrated.
lntegration costs can be expressed altematively in terms of incremental costs. Integration costs
when expressed incrementally assume early projects are assessed lesser integration costs,
and later projects need to make up the difference to allow full cost recovery for a given build-out
scenario. For example, if solar plants comprising the first 100 MW build-out are assessed
integration costs of $0.40iMWh, then plants comprising the increment between 100 MW and
300 MW need assessed integration costs of $l.50iMWh to allow full recovery of the $1.204/twh
costs to integrate 300 MW of solar plant capacity. lncremental solar integration costs are
provided in Table 9.
Table 9
lncremental integration cost results for solar build-out scenarios
0-100 Mw t00-300 Mw 300-500 Mw s00-700 Mw
lncrcmental integration cost $0.40/l\llVVh $1.50/MWh $2.8o/MWh $4.40/t\Mh
Note: Costs are in 2014 dollars and rounded from simulation results to the nearest $0.10.
Page 15
Solar lntegration Study Report ldaho Power Company
Study Findings
Hour-Ahead Sola r Production Forecasting
Analyses suggest a persistence-based forecast with adjustrnent to account for known changes
in the sun's position provides a reasonable production forecast for hour-ahead operations
scheduling. The persistence-based hour-ahead solar production forecast used for the study is
based entirely on observed production and consequently could be readily adopted in practice.
While a day-ahead solar production forecast would be necessary in practice for a balancing
authority integrating solar, deviations from the day-ahead forecast can be managed through a
combination of market transactions and operations modifications, and consequently the study
imposes no reserve requirement to cover deviations for day-ahead solar production forecasts.
Compared to wind, system operators managing a balancing authority integrating solar would
have the benefit of at least six hours at the start of day with no or little solar production.
During this period of no or little solar production, system operators could evaluate the day-ahead
solar production forecast using information from updated weather forecast products and begin to
plan for necessary actions to manage deviations from the day-ahead solar production forecast.
In contrast, deviations from the hour-ahead solar production forecast can only be covered by
Idaho Power's dispatchable generators. The analysis for the solar integration study by design
determines the amounts of bidirectional capacity in reserve that system operators would need to
schedule to position dispatchable generators to cover possible deviations from the hour-ahead
solar production forecast. Integration costs are a result of the sub-optimal scheduling of the
dispatchable generators associated with holding the solar-caused capacity in reserve.
Comparison to Wind lntegration
This study indicates solar plant integration costs are lower than wind plant integration costs.
The lower integration costs associated with solar are fundamentally the result of less variability
and uncertainty. As described in the preceding section, the study assumes deviations in solar
plant production from day-ahead forecast levels can be managed through a combination of
market transactions and operations modifications, allowing day-ahead generation scheduling to
avoid extra reserye burden. Therefore, reserves carried for solar generation can be focused on
readying dispatchable generators to respond to unplanned solar excursions from hour-ahead
production forecasts. Moreover, logic incorporated in the derivation of lower and upper bounds
on the hour-ahead production forecast, which can be readily adopted in practice, allows the
adjustrnent of the bounds in response to observed solar production patterns. ln effect,
the hour-ahead forecast is based on a persistence oflevel ofproduction (adjusted for the
known change in the sun's position), as well as a persistence of variability in production.
The consequence of these methods is that bidirectional capacity held in reserve on dispatchable
generators to respond to solar variability and uncertainty is less than that required for responding
to wind.
Qualitatively, solar is more predictable than wind. Sunrise and sunset times, as well as the
time of solar noon, are a certainty. The theoretical maximum level of production can be
Page 16
ldaho Power Company Solar lntegration Study Report
readily derived, reflecting patterns on daily, monthly, and seasonal time scales. Finally,
land requfuements for a solar power plant are likely to promote a relatively high level of
dispersion, which is critical to the mitigation of impacts from severe and abrupt ramps in
production exhibited by individual panels in response to passing clouds. The effects of
geographic dispersion are discussed further in the following section.
Geographic Dispersion
Production for a single solar photovoltaic panel exhibits severe and abrupt intermittency
during variably cloudy conditions; a TRC member expressed during a meeting that for a single
panel, the drop in production from a cloud is effectively instantaneous. The effect of severe and
abrupt intermittency is commonly attributed to the absence of inertia in the photovoltaic process.
While the intermittency effect is severe for a single panel, dampening occurs when considering
the production from a solar plant-sized aggregation of panels, and even further dampening occurs
when considering the production from several solar plants spread over a region such as southern
Idaho. Therefore, geographic dispersion has significant influence on solar integration impacts
and is perhaps of greater importance for solar than wind.
The four studied solar build-out scenarios each have capacity installed at six southem Idaho
locations spread over more than 220 miles from east to west. Because of the substantial
geographic dispersion, severe instantaneous ramps in solar production for the study data are
relatively infrequent. If solar plant development in southem ldaho occurs in a more clustered
fashion than assumed for this study, actual integration impacts and costs will be higher than the
results of this study.
Transmission and Distribution
The focus of tdaho Power's solar integration study is a macro-level investigation of the
operations modifications necessary to maintain balance between power supply and customer
demand for a balancing authority integrating photovoltaic solar plant generation. The objective is
to understand the impacts and costs of the sub-optimal operation of dispatchable generating
capacity. The study is not an investigation of integration issues related to the delivery of energy
from proposed solar photovoltaic power plants to the retail customer; these issues are addressed
in individual interconnection studies performed on a plant-by-plant basis.
Spring-Season I nteg rati on
The production cost simulations suggest reserve requirements are particularly problematic when
hydroelectric resources are highly constrained, such as frequently occurs during spring-season
periods characterized by high water, low customer demand, and high generation from variable
generating resources, such as wind and solar. Experience has shown wind integration to be
particularly challenging during these periods, and the simulations suggest similar challenges
integrating solar. This study finding is corroborated by NREL in the Western Wind and Solar
Integration Study Phase 2 (Lew et al. 2013), which reports the need for flexibility is notably high
during the spring and that during these periods the curtailment of variable generation is one
source of flexibility enabling the balancing of generation and customer demand.
Page 17
Solar lntegration Study Report ldaho Power Company
Cottcr-usroNS
The cost to integrate the variable and uncertain delivery of energy from solar photovoltaic power
plants is driven by the need to carry extra capacity in reserve. This extra capacity in reserve is
necessary to allow bidirectional response from dispatchable generators to unplanned excursions
in solar production. The simulations performed for Idaho Power's solar integration study indicate
the costs associated with holding the extra capacity in reserve (Table 8).
Further Study
The integration of variable generation, including the study of methods for determining
integration impacts and costs, continues to be the subject of considerable research. The breadth
of this research highlights the interest in variable-generation integration, as well as the evolution
of study methods. Idaho Power appreciates the level of interest in its study of integration of
variable generation and recognizes the likelihood of a second-phase study with expanded scope.
During the course of the solar integration study, in discussions with the TRC and participants of
the public workshop, Idaho Power has received suggestions for a second-phase study of solar
integration. Suggestions for a second phase include the study of the following:
o Altemative water-year types (e.g., low-t1,pe and high-type)
o Intra-hourtradingopportunities
o Shortening the hour-ahead forecast lead time from 45 minutes to 30 minutes
o Clustered solar build-out scenarios
o Smaller solar build-out scenarios (e.g., 50 MW of installed capacity)
o Other solar plant technologies (e.9., tracking systems or varied fixed-panel orientation)
o Distributed solar systems (i.e., rooftop systems)
o Correlation between solar, wind, and load variability and uncertainty
o Improved forecasting methods
. Energy imbalance markets
o Voltage/frequencyregulation
Idaho Power will consider these suggestions during the development of scope for a
second-phase study.
Page 18
ldaho Power Company Solar lntegration Study Report
LrenRrURE Greo
Ela E., M. Milligan, B. Parsons, D. Lew, and D. Corbus. 2009. The evolution of wind power
integration studies: Past, present, and future. IEEE Power & Energy Society General
Meeting,2009 PES '09.
Idaho Power.2013. Wind Integration Study Report. Boise, ID: Idaho Power.
Lave, M., A. Ellis, and J. Stein. 2013a. Simulating solar power plant variability:
A review of current methods. Sandia report-SAND2013-4757.
http://enerey. sandia. eov/wp/wp-contenUeallery/uploads/SAND20 I 3 -
4757_Simulatine Solar_Power Plant Variability A_Review of Current_Methods_Fl
NAL.pdf. Accessed June 2014.
Lave, M., J. Kleissl, J. S. Stein. 2013b. A wavelet-based variability model (W\IM) for solar PV
power plants. IEEE Transactions on Sustainable Energy, Vol. 4, No. 2.
Lew, D., G. Brinkman, A. Ibanez, A. Florita, M. Heaney, B. M. Hodge, M. Hummon, G. Stark,
J. King, S. A. Lefton, N. Kumar, D. Agan, G. Jordan, and S. Venkataraman.2013.
Westem Wind and Solar Integration Study Phase 2. NREL/TP-5500-55588. Golden, CO:
National Renewable Energy Laboratory. http://www.nrel.gov/docs/ff13osti/55588.pdf.
Accessed June 2014.
ILIVIG and NREL]. Utility Variable-Generation Integration Group and National Renewable
Energy Laboratory. No date. Principles forTechnical Review Committee (TRC)
involvement in studies of variable generation integration into electric power systems.
http://variablegen.org/wp-contenVuploadsl2009l05/TRCPrincipleslune2012.pdf.
Accessed June 2014.
Page 19
Solar lntegration Study Report ldaho Power Company
This page left blank intentionally.
ldaho Power Company Solar lntegration Study Report
Appendix 1
Solar integration study appendix
Table of Contents
Introduction
Technical Review Committee
List of TRC Members
Regulatory Commission Staff Observers
TRC Schedule and Agenda
Public Workshop Schedule and Agenda
Data Inputs and Assumptions
Natural Gas Price Assumptions
Market Power Price Assumptions
IPC Customer Load Data
Idaho Power Existing Generation
Hydroelectric Generation Data
Run-of-River Projects
Wind Generation Data
Aggregate PPA and PURPA Projects
Non-Wind PURPA Generation Data
Solar Production Data
Derivation of Hour-Ahead Solar Production Forecast and Upper/Lower Bounds
Page 21
Solar lntegration Study Report ldaho Power Company
lrurRooucnoN
This appendix contains supporting data and explanatory materials used to develop Idaho Power's
2014 Solar Integration Study.
The main document, the 2014 Solar Integration Study, contains a full narrative of Idaho Power's
process for studying solar integration costs. For information or questions concerning the study,
contact Idaho Power:
Idaho Power-Resource Planning
l22l W.ldaho St.
Boise,Idaho 83702
208-388-2623
Tecn rurcAL Revrew Gorvuu rree
The Technical Review Committee (TRC) was formed during surlmer 2013 to provide input,
review, and guidance for the study. It is comprised of participants from outside of Idaho Power
that have an interest and/or expertise with the integration of intermittent resources onto
utility systems.
As part of preparing the 2014 Solar Integration Study,Idaho Power held one public meeting and
four TRC meetings. Idaho Power values these opportunities to convene, and the TRC members
have made significant contributions to this plan.
List of TRC Members
Brian Johnson...................University of Idaho
Jimmy Lindsay.................Portland General Electric (formerly of Renewable Northwest Project)
Kurt Myers ....Idaho National Laboratory
Paul Woods ......................(formerly of City of Boise)
Cameron Yourkowski......Renewable Northwest Project (replacing Jimmy Lindsay)
Regulatory Commission Staff Obseruers
Brittany Andrus................Public Utility Commission of Oregon (OPUC) staff
John Crider ....OPUC Staff
Rick Sterling....................Idaho Public Utilities Commission (IPUC) staff
ldaho Power Company Solar lntegration Study Report
TRC Schedule and Agenda
Meeting Dates Agenda Items
2013 Thursday, August 15 lntroductions and role of TRC
ldaho Power system overview
Formulation of basic study design
Establish solar futures
Techniques for building solar generation data
Closing thoughts and comments
2013 Thursday, September 19 Study design
Key study components
Hydro-WY 201 1 vs. WY 2012 vs. WY 2013
Solar-WY 2011 vs. WY 2012 vs. WY 2013
Market power prices
Natural gas prices
Solar penetration levels
2014 Monday, January 6 Review of Study Design
Solar Data Availability
Wavelet-based Variability Model
Analysis Conclusions
2014 Friday, May 16 Review of lntegration Study Design
Review of IPUC Filing
Development of Reserve Requirement for solar scenarios
2014 Thursday, May 29 Review of Operating Reserves
Review of Production Cost Model
Public Workshop Schedule and Agenda
Meeting Dates Agenda ltems
2014 Thursday, May 1 lntroduction of Technical Review Committee
ldaho Power system overview
Study objective
Study design
System modeling
Next steps
Page 23
Solar lntegration Study Report ldaho Power Company
Natural Gas Price Assumptions
Table 1
Actual monthly average Nymex price for water year 2012
Average Monthly Price
Dara lnpurs AND AssulvtproNs
$3.76
$3.52
$3.36
$3.08
$2.68
$2.45
$2.19
$2.04
$2.43
$2.77
$3.01
$2.63
$26.02
$30.81
$30.13
$24.s3
$23.50
$16.30
$8.9e
$5.81
$4.50
$12.0s
$24.75
$24.47
2011
2012
October
November
December
January
February
March
April
May
June
July
August
September
October
November
December
January
February
March
April
May
June
July
August
September
Market Power Price Assumptions
Table 2
Actual average Mid-Columbia dollars/megawatt-hour (MWh) for water year 2012
Average Monthly Price
2011
2012
Page 24
ldaho Power Company Solar lntegration Study Report
IPC Customer Load Data
Table 3
Actual average megawatt (MW) for waler year 2012
Year Month Average Load
2011 1,403
1,s63
1,729
1,680
1,s97
1,457
1,504
1,742
2,108
2,388
2,197
'l,679
2012
Hydroelectric Facllities and
ilamepLte Capadtles
I xells ca;rron 391.5 Mw
190.0 MW
585.4MW
12.4 MW
27.2 MW
82.8 MW
75.0 MW
13.5 MW
8.3 Mtr'
@ tow€rsalmon 60.0MW
E Upp€rsalmm 34.5Mw
Elhousandsprings 8.8Mw
OREGOI{
E oeartalc 2.5 MW
E Sho.hon€Falls 12.5Mw
E Twin talh
El Milner
E American Falls ??:lMw
Iord 1,709.0 tt !
^t{orth \hlmy
NEVADA
Figure 1
Existing ldaho Power generating resources
Therm.l F.cllltlcs And C.p.cttles
Coal
A Jim B.i4er 770.5 MW'
A l{sth lblmy 283.5 MW'
A Eoadman _,91!iry,'
ror.l _1|1lqj? !ll!v
t.rtrrd G8
A B€nnettMountain l72.8Mw
A Danstin 270.9MW
A Lantleyculch !!8.5MW
rot l ,62.2 MW
Ole3el Salmon Diesel 5.0 MW
tor.l 1,885.4 MW
WYOMIITIG
October
November
December
January
February
March
April
May
June
July
August
September
ldaho Power Existing Generation
E oxlow
! arownlee
E c.$ade
E Swan talls
E c l. str*e
E uiss
E Lourer Mahd
E UppeMalad
WASHII{GTON
fiEJP
s2.9 MW
59.1MW
Page 25
Solar lntegration Study Report ldaho Power Company
aa
E zs.o?.,
I.l
E 2o.o
=E.E r's.o
=o
EE 1o.oato
ot s.o.
1951-2013 avcraSc
a
oo.)l
o.o
1949
Figure 2
Brownlee Reservoir inflow by water year
447
4',t8
4'.t5
358
365
380
388
252
337
292
251
208
February
March
Apill
May
June
July
Hydroelectric Generation Data
Run-of-River Proiects
Table 4
Actual monthly average MW (aMW) for water year 2012
2011 October
November
December
January2012
August
September
Page 26
ldaho Power Company Solar lntegration Study Report
Aggregate PPA and PURPA Projects
Table 5
Actual monthly aMW for water year 2013
2011
2012
Non-Wind PURPA Generation
Table 6
Actual monthly aMW for water year 2012
Wind Generation Data
October
November
December
January
February
March
April
May
June
July
August
September
95
190
120
194
167
191
172
166
163
144
131
116
Data
2011
2012
October
November
December
January
February
March
April
May
June
July
August
September
96
52
45
43
43
54
104
135
131
140
130
'111
Page27
Solar lntegration Study Report ldaho Power Company
Solar Production Data
Disper3ed 100 ilW: Soasonal Aveiage Daily Shapo
rm
90 -
SmmsAvsage Ju, Jd, AW
-SpnngAsAq
Mz, AW,W
80
-70o
3*
5
t*iz3+oI{l30
m
10
0
0
- - Amrd AwAe: Oct- S€p
"-"". FdlAvsee: S€p, Oct, tlov
.-- WirtsAveree: D6, Ja, Feb
00 l:m 2:00 3:00 4:00 5:00 6:00 7:00 8i00 9:00 10:00 '11:00 12:m 13:00
Tlm ol Day
."""t:.;":$511;:.5i5!$5:Xlgl.''.""""".""s
Figure 3
Dispersed 100 MW
,"$}:l;'"'
Dispersed 100 MW-Productlon
Page 28
ldaho Power Company Solar lntegration Study Report
l
l
l
300
2fi
2@
24
m
9r*I3 raoE
5 160c
t ',+o!
f,ro
.fI,*
80
40
m
0
0 1:00 2:00 3:00 4:00 5:00 6:00 7:00
Oispeced 300 MW: Seasonal Avorags Daily Shapo
-SmmsAvsage
Jm, Jd, Aug
-SpringAvq{6:
Mil, Apr, May
8:00 9:m 10:m 11:m 12:@ 13.00 14im 1100 16:00 17:m lAm 19m 20100 21:m 22:@ 2}@
Tlmof Day
Dispe.sed 300 iilw-Production
90
80
70
60
50
40
30
20
10
0
Figure 4
Dispersed 300 MW
o
3.
""Ii
jl.:-1":"":t1"*oo":.:".-}:}',.
Page 29
Solar lntegration Study Report ldaho Power Company
Oisp€rsed 500 MW: Seasonal Average Daily Shape
500
450
400
_ 350o
=*3!$r*oi
2zmtIo- tso
1m
-
S0msAvsagerJm, Jd, ArB
-Spring
Avseei Mar, Apr, May
- - Am€lAwrage:Oct- Sep
FallAvdago: S€p, Ocl, Nov
- - V\4rtsAsage D@, Ja, Feb
0r
i
I
0100 1:00 2:00 3100 4:m 5:00 6:00 7:00 8:00 9:00 10:m 11:m 12:00 13:00 t4:m 15:m ,6:m 17:m 18rm 1900 20:00 2t:00 2,:n 2}n
Tlm of Oa!,
Disperced 500 Mw-Production
l
.C.,""" o,9 ,S"ooot'.d'o.doo ,"." .""" C """
"so"sd".s$asdao"""".r"s".rr'
-$ -$ -'"o -\s -Sa.C ".c" -".do ".d"" a.dt
"f -.S -\s
"--1$si""."1.."s
F.-
Figure 5
Dispersed 500 MW
o
==I
I 4'l
Page 30
f'"."$$"t'
ldaho Power Company Solar lntegration Study Report
Dispersed 700 MW: Sea$nalAvtrage Daily Shape
7m
650
6m
550
500
o
Io*.
Eo*
$ gsoo
;*!l rrotN
150
1m
50
0
-l
-
Smms Av6age: J0, Jrl, Aug ]
llrl
-
Sp.irE Avtr4e: Ma, Apr, May
Ill
- - Amul Avsage: Oct- Sep il
i
IFallAvsage: Sep, Oct, Nov i
I
WintsAvtrage:D6, Jil, Feb I
l
6:00 7:00 8:00 9:00 10:00 11:00 12:00 1300 14rm 15:00 16:00 17:00 18rm 1gm 2A0o 21.N 22N 23tc[
Tlmof Day
"..::1.:..1.:i.51ti..:1.:_...$1I1X91.,""""..,"."..d
Figure 6
Dispersed 700 MW
-$ -s -*d -..f
""1isi"-'-'t"""c
Page 31
Solar lntegration Study Report ldaho Power Company
Derivation of Hour-Ahead Solar Production Forecast and
Upper/Lower Bounds
95
90
85
80
75
70
9.'
Euo
.E*
E'o
Eot0,940
!3soul 30
20
15
10
. AdJl@WOdF
-[ Dry h* k., da 'tt tu.@.
. brdlhrkil
. kor^h.dtu.-tA[
. hr.ci UpFr &und
6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00
Figure 8
Hour-ahead forecast example
The average forecast is shown on Figure 8 as the green series. For each hour of the day, the
forecast average is calculated by applying the follow equation:
Forecast AugG) : Forecast Obs(MW)(r-r,oo- r_1:15) *
Where:
I : forecast hour
Aug CSI56:oo+ r:55)
Av g C S I S g_zilo+r_1:15)
C^S/S: Clear Sky lndex Surrogate
The Clear Sky Index Surrogate (CSN) is an important measure of the maximum amount of solar
generation the system could experience in any given hour. The CSIS is a component of the
average solar production forecast and accounts for the seasonal changes that influence solar
photovoltaic generation. This value is unique for every hour of the year. The CSIS is calculated
using 5-minute, modeled production data from the wavelet-based variability model (WVM).
The CSIS is calculated by taking the maximum S-minute observation for a given hour.
This maximum value is the absolute maximum for a given hour over a lO-day period.
After identifying the absolute maximum from water year 2011, the forecast also identifies the
absolute maxima for water years2012 and 2013. With the three absolute maxima identified from
the three water years analyzed, the forecast applies the maximum CSIS observed in three years
0
5:00
7l8l2OL2
Page 32
ldaho Power Company Solar lntegration Study Report
of data for a given hour. It is noted that the ratio of the CSIS values, described in the above
equation, result in the least amount of average production forecast error. Multiple variations of
this ratio were tested, and the final version of the ratio was the most accurate. The process
detailing the calculation of the CSIS is described in the equations below.
CSIS(1; = L4ax ([cslslw**r ear2ort)], [cstslr*- yeu2072)f,lcstsg**r,,,rorr)])
Where:
cslslweteryeorzott'1 = uax (fsminobs(MW)u,,,-r],[srinoatl,uw)a>ro-rf,...,[s*inobs(Ml{r)1a,o-,",])
cslSlwatertear*o.2) = uax ([sminobs(MW)u,,r-r],[srin obs(MW)1g1,r-r],...,[srninoat{ruw).r,0-,r])
cslslwttcrvcorzorr) = uot ([sminobs(MW)u,,r-r],[srin lbs(MW)1r1ro-r],..,[srninoat{uw),r,o--,])
Where:
f= forecast hour
d= forecast day
Figure 8 is a good example of how the persistence-based forecast does very well under the
majority of solar conditions and how a forecasting model struggles with extreme weather events.
Despite the limitations of a persistence forecast, within a short period of time the forecast
retumed to accurate predictions. Figure 8 is a select, extremely variable generation profile.
The aftemoon observations that fall beneath the lower bound forecast are included in the
2.5 percent of lower forecast error reported in the solar integration study. Generally, the forecast
does well capturing the variability in production due to solar. The forecast has the ability to
tighten the range between the upper and lower bounds. This ensures the amount of capacity held
in reserve is sufficient but not unduly large.
Page 33
Solar lntegration Study Report ldaho Power Company
This page left blank intentionally.