TEST SYSTEM

CSP was implemented in a test system to better verify the basic perfor­mance of this dispatchable energy source and to more easily isolate the relative value of TES under various scenarios.

The best locations for CSP in the United States are in the desert south­west within the Western Interconnection. Simulating the entire intercon­nection makes it difficult to isolate the performance of CSP, so a smaller test system was created to develop and validate the modeling approach.

image179

FIGURE 5: Load duration curve in 2020 for the PSCO/WACM test system

Most of the existing and proposed CSP is in California; however, simply running the California system in isolation ignores the substantial intercon­nections between the California and bordering states. As an alternative, we developed a system composed of two balancing areas largely in the State of Colorado: Public Service of Colorado (PSCO) and Western Area Colorado Missouri (WACM). These balancing areas consist of multiple individual utilities and this combined area is relatively isolated from the rest of the Western Interconnection. In addition, Colorado has sufficient solar resource for CSP deployment in the San Luis Valley in the south­central part of the state, and there have been proposals for large-scale solar development in the area (Xcel 2011). The test system also has sufficient wind resources for large-scale deployment, which makes evaluation of high renewable scenarios more realistic.

The Colorado test system was isolated by physically “turning off” the generation and load and aggregating the transmission outside of the PSCO and WACM balancing areas. Transmission was modeled zonally, without transmission limits within each balancing authority area. It is very difficult to simulate any individual or group of balancing authority areas as actually operated because the modeled system is comprised of vertically integrated utilities that balance their system with their own generation and bilateral transactions with their neighbors that are confidential. Not having access to that information, we modeled the test system assuming least-cost eco­nomic dispatch. We based our inputs and assumptions as much as possible on the Western Electricity Coordinating Council (WECC) Transmission Expansion Policy Planning Committee (TEPPC) model and other publicly available datasets. Projected generation and loads were derived from the TEPPC 2020 scenario (TEPPC 2011). Hourly load profiles were based on 2006 data and scaled to match the projected TEPPC 2020 annual load. The system is a strongly summer peaking system with a 2020 coincident peak demand of 13.7 GW and annual demand of 79.0 TWh. The system load duration curve for 2020 is shown in Figure 5.

The generation dataset was derived from the TEPPC 2020 database and included plant capacities, heat rates, outage rates (planned and forced), and several operational parameters, such as ramp rates. A total of 201 ther­mal and hydro generators were included in the test system, with total ca­pacities listed in Table 2. The generator database was modified to include part-load heat rates based on Brinkman et al. (2012). Start-up costs were added using the start-up fuel requirements in the generator database plus the operations and maintenance (O&M) related costs based on estimates prepared for the WWSIS II study (Intertek/APTECH 2012). We adjusted the generator mix to achieve a generator planning reserve margin of 15% by adding a total of 1,450 MW (690 MW of combustion turbines and 760 MW of combined cycle units).

TABLE 1: Characteristics of the Test System Conventional Generators in 2020

System Capacity (MW)

Coal

6,178

Combined Cycle (CC)

3,724

Gas Turbine/Gas Steam

4,045

Hydro

773

Pumped Storage

560

Other"

513

Total

15,793

1Includes oil and gas-fired internal combustion generators and demand response.

Two renewable energy scenarios were created by adding wind and solar generation. PV profiles were generated using the SAM model with 2006 meteorology. Wind data was derived from the WWSIS dataset. A low renewable energy (RE) case was created by adding wind and solar to achieve a penetration of about 13% on an energy basis. This is a relatively small increase over the renewable penetration in 2011; Colorado received about 12% of its electricity from wind in the year ending June 2012 (EIA

2012) . We also considered a high RE case where wind and solar provide about 35% of the region’s energy. In each case, discrete wind and solar plants were added from the WWSIS data sets until the installed capacity produced the targeted energy penetration. The sites were chosen largely based on capacity factor, and do not necessarily reflect existing or planned locations for wind and solar plants. Table 2 lists characteristics of the sys­tem in the two cases, while Table 3 provides additional details of the re­newable and conventional generation mix.

TABLE 2: Renewable Scenarios in the Test System in 2020

System Capacity (MW)

Low RE Scenario

High RE Scenarioa

Wind Capacity (MW)

3,054

6,489

Wind Energy (GWh)

9,791

20,210

Solar Capacity (MW)

395

3,630

Solar Energy (GWh)

625

6,493

a This is the potential generation and does not include curtailment that results in actual dispatch. About 31 GWh of wind and 23 GWh of solar were curtailed in the base high RE scenario.

Three classes of ancillary service requirements were included. The contingency reserve is 810 MW based on the single largest unit (Coman­che 3). This reserve is allocated with 451 MW to PSCO and 359 MW to WACM, with 50% met by spinning units. Regulation and flexibility reserve requirements were calculated based on the statistical variability of net load described by Ibanez et al. (2012). Reserves were modeled as “soft constraints,” meaning the system was allowed to not meet reserves if the cost exceeded $4,000/MWh. This high cost could result during peri­ods where a power plant would need to start up for a very short period of time just to provide reserves. Load was also modeled as a soft constraint, with a loss-of-load cost of $6,000/MWh (though the reserve margin was adequate to avoid lost load).

Fuel prices were derived from the TEPPC 2020 database. Coal prices were $1.42/MMBTU for all plants. Natural gas prices varied by plant, and for most plants were in the range of $3.9/MMBTU to $4.2/MMBTU, with a generation weighted average of $4.1/MMBTU. This is slightly lower than the EIA’s 2012 Annual Energy Outlook projection for the delivered price of natural gas to the electric power sector in the Rocky Mountain region of $4.46/MMBTU in 2020 (EIA 2012). Sensitivity to natural gas price was also analyzed.

TABLE 3: Base Case Results

Low RE

High RE

Total Production Cost (M$)

1,491.37

1,024.38

Average Production Cost ($/MWh)

18.9

13.0

Total Generation (GWH)’

78,957

79,098

Generation Mix

Coal

58.8%

52.0%

Gas Combined Cycle (CC)

20.7%

7.2%

Gas Combustion Turbine

(CT)/Gas Steam

1.4%

1.1%

Hydro

4.8%

4.8%

Wind

12.4%

25.5%

Solar PV

0.8%

8.2%

Other

1.1%

1.2%

Fuel Use (1,000 MMBTU)

Coal

490,923

434,426

Gas

140,447

53,928

aWhile the load is the same, the total generation is slightly different (by about 0.2%), due primarily to different operation of the pumped hydro units.

Both cases were run for 1 full year (2020, with 2006 meteorology and load pattern). The model run begins with two scheduling models to de­termine outage scheduling and allocate certain limited energy resources. The model then performs a chronological hourly security-constrained unit commitment and economic dispatch to minimize the overall production cost under operational and system constraints. The model performs a 24- hour ahead commitment with an additional 24-hour look-ahead period, al­lowing the model to effectively optimize storage utilization over a 48-hour period. The analysis in this report was performed using PLEXOS version 6.207 R01, using the Xpress-MP 23.01.05 solver, with the model perfor­mance relative gap set to 0.5%.

Table 3 provides a summary of the operational results for the two base simulations. This represents on the variable cost of system operation, dominated by the cost fuel for thermal power plants. There was no loss of load and a small number of reserve violations (less than 40 hours per year in both cases).

image180

FIGURE 6. Dispatch stack during the period of July 25-28 in the low RE Case

image181

FIGURE 7: Dispatch stack during the period of February 8-11 in the high RE Case

System dispatch stacks can provide additional insight into system op­eration. The generation mix and dispatch was as expected, with coal units operating as baseload units, and CC and CT units operating as mid-merit and peaking units as needed. Figure 6 shows the dispatch stack for the low RE case during the week of peak demand in the summer. This figure dem­onstrates the opportunity for mid-day solar generation to reduce the use of the highest cost generators.

The high RE case removes much of the gas generation from the system and leaves coal on the margin for a large number of hours. Figure 7 shows a four-day period in February, which includes the day of the lowest net demand on the system in the high RE scenario. In the first two days, coal generators reduce their output to minimum levels and renewable genera­tion is curtailed. Any additional renewable generation from 1 p. m. to 3 p. m. on February 8 and from 11 a. m. to 2 p. m. on February 9 will be unus­able in this scenario, and likely curtailed. Generation during many other periods will offset mainly lower-cost coal generation. Also of note is the rapid increase in net demand that occurs after 3 p. m. when the decrease

image182

FIGURE 8: System marginal price duration curve in the PSCO balancing area for the two cases

in solar output and increase in electricity demand require large ramps of the coal units, use of higher-cost combustion turbines, and dispatch of the pumped storage plants. Previous integration studies such as WWSIS have found significant increases in ramping requirements of coal units, and a major focus of the second phase of WWSIS is to examine the potential cost implications of increased unit cycling.

In any hour of the year, the value of solar or other incremental genera­tion in this system is determined by the marginal generators and associ­ated price. Figure 8 is a price duration curve for the test system showing system marginal cost for the PSCO balancing area. The marginal prices for the WACM balancing area were almost identical because transmis­sion constraints between the PSCO and WACM system were not binding; only very small price differences occurred in a few hours due to different reserve requirements. The price duration curve shows three main “zones” of prices, based on the marginal generators: coal at about $17-$20/MWh, combined cycle units at about $25-$35/MWh, and combustion turbines at about $38-$45/MWh. In the low RE case, coal is on the margin about

for about 1,600 hours, while in the high RE case, coal is on the margin for over 5,000 hours. In addition, renewable energy is effectively on the mar­gin for about 100 hours of the year in the high RE case; additional renew­able generation during these hours would likely be curtailed and provides no incremental benefit to the system. There are also a small number (less than 40 hours per year) of extremely high prices, set by the reserve viola­tion conditions.