An Analysis of Concentrating Solar Power with Thermal Energy Storage in a California 33% Renewable Scenario
Paul Denholm, Yih-Huei Wan, Marissa Hummon, and Mark Mehos, March 2013 (National Renewable Energy Laboratory)
Concentrating solar power (CSP) with thermal energy storage (TES) is a dispatchable source of renewable electricity generation. However, the dispatchability of this resource is limited by the availability of solar energy. This makes it challenging to quantify the value of CSP and provide comparisons to alternative generation sources.
The California Independent System Operator (CAISO) has prepared a number of analyses of the grid operational challenges associated with the state’s 33% renewable portfolio standard (RPS). These analyses, which used a commercial production cost model, created a publically available database of the CAISO system. This database can be used as a basis for analyzing the potential value of CSP with TES in California.
This analysis used the “Environmentally Constrained” 33% RPS scenario database in the
PLEXOS grid simulation tool to estimate the value of CSP in avoiding conventional fossil generation, and compared this value to other sources of generation. To perform this analysis, we created a baseline scenario and added four types of generators, each in a separate scenario. The four generator types were photovoltaic (PV), a baseload generator with constant output, a CSP plant providing dispatchable energy, and a CSP plant providing both energy and operating reserves. Each generator added the same amount of energy (about 1% of annual demand) for an equal comparison of their value. In each case, we calculated the difference in production costs between the base case and the case with the added generator. This difference in cost was attributed to the added generator as its operational value to the system.
PLEXOS dispatches the hourly energy inflow of solar energy in the CSP plant to minimize the overall system production cost. The model considers the interaction of the California system with the rest of the Western Interconnection, and new generators within California can therefore affect the dispatch of coal, gas, and other generators throughout the West.
The operational value of each generator is associated with avoided fuel (and associated
emissions) as well as reduced operations and maintenance (O&M) and power plant start costs. In addition to operational value, generators add capacity value to the system that can be estimated by examining generator operation during periods of high net demand. The CSP plants in this study provided energy during essentially all high net demand hours, implying a capacity credit similar to a conventional dispatchable resource. The corresponding value is typically determined by a proxy resource, such as a combustion turbine, or alternative market-based mechanism. In this analysis, we use a low capacity cost of $55/kW-yr and a high cost of $212/kW-yr.
Table ES-1 Summarizes the value estimated by combining the operational results from the PLEXOS simulations with capacity value estimates for each technology.
Overall, the analysis demonstrates several properties of dispatchable CSP including the
flexibility to generate during periods of high value and avoid generation during periods of lower value. Of note in this analysis is the fact that significant amount of operational value is derived from the provision of reserves in the case where CSP is allowed to provide these services, adding about $17/MWh. This represents a substantial change in operational practice, including frequent operation at part-load. The incremental value of CSP with TES in this scenario was $30/MWh to $51/MWh compared to a baseload resource, or $32/MWh to $40/MWh compared to PV. This range depends on both the ability of CSP to provide operating reserves and the expected cost of new capacity.
This analysis also indicatesthat the “optimal” configuration of CSP may vary as a function of renewable penetration, and each configuration will need to be evaluated in terms ofits ability to provide dispatchable energy, reserves, and firm capacity. As the net load variability increases with more renewable generation, CSP plants with different solar field sizes, amount of storage, and ramp flexibility may be best suited to enable integration of these variable-generation (VG) resources. This will also change the value proposition for CSP with TES. Future analysis will consider these elements under alternative RPS scenarios, including higher fractions of energy derived from renewable resources.
In summary, NREL has implemented a methodology for evaluating the operational impacts of CSP systems with TES within the PLEXOS production cost model. This model was used to quantify the additional value provided by this flexible resource as compared to baseload or VG resources. The model can be used to investigate additional scenarios involving alternative technology options and generation mixes, applying these scenarios within California or in other regions ofinterest.
The California Independent System Operator (CAISO) has performed a number of analyses of the grid impacts and operability of various scenarios associated with meeting California’s 33% renewable portfolio standard (RPS) in 2020 (CAISO 2011a). These scenarios, developed by the California Public Utilities Commission (CPUC), have included various amounts of concentrating solar power (CSP). However, to date, the CSP plants in the scenarios evaluated by the CAISO have not included significant amounts of dispatchable thermal storage. As of early 2013, contracts with such plants have been approved and there is increasing interest in the potential benefits offered by plants deployed with thermal energy storage (TES). CSP with TES is a dispatchable source of renewable energy and can provide valuable grid flexibility services including the ability to shift energy in time and provide both firm capacity and ancillary services. This flexibility can potentially aid in integrating variable-generation (VG)sources such as
photovoltaics (PV) and wind and further reduce the overall production cost in a system when compared to a renewable portfolio of equal energy but without TES.
This document describes a preliminary evaluation of CSP with TES in the CAISO system, based on one of the scenarios developed for the 33% RPS study.CSP with TES was incorporated into the PLEXOS production cost model, and the differences in production cost were analyzed for the CPUC’s “Environmentally Constrained” scenario. Specifically, the incremental value of CSP with TES providing about 1% of CAISO demand was evaluated, and was also compared to PV and a baseload resource providing the same amount of energy. It should be noted that this work does not evaluate any of the capital or operational costs of any of the technologies evaluated.
Overall, the analysis demonstrates several properties of dispatchable CSP including its ability to generate during periods of high value and avoid generation during periods of lower value. Of note in this preliminary analysis is that significant operational value is derived from providing ancillary services that require frequent operation at part-load…
Results…Overview of CSP Operation…
The qualitative overview of CSP operation presented in section 3.1 can be translated into the actual impact on system production costs. The operational value of each technology represents its ability to avoid the variable cost of system operations using the resource mix assumed in the scenario. The CAISO model tracks operational costs in four cost categories—operating fuel, variable O&M, start-up costs, and emissions costs. Operating fuel includes all fuel used to operate the power plant fleet while generating and includes the impact of variable heat rates and operating plants at part load to provide ancillary services. It does not include any penalties associated with loss of load or reserve violations, because there was no shortage of energy or reserves in any of the simulations.
Table 3 summarizes the results from the production simulations. It provides value per unit of energy delivered, calculated by dividing the difference in production cost by the total energy delivered to the grid by each technology.
Table 3 demonstrates a relatively small increase in the operational value of an energy-only CSP plant compared to a baseload resource (about $6/MWh) or a PV plant (about $12/MWh), but a much greater difference when the CSP plant is able to provide reserves. Adding the ability to provide reserves increases the operational benefits of CSP by about $17/MWh, or a difference of about $22/MWh compared to the baseload resource and about $29/MWh compared to PV. A large fraction of the difference between the CSP plant with reserves and PV is the cost of starts, with PV increasing the net variability and reserve requirements, which increase the number of thermal plant starts. We assume CSP does not add to system reserve requirements and displaces thermal unit starts when providing energy, ramping, and providing reserves.
As discussed previously, several simplifying assumptions were made when simulating CSP, and these could have an impact on the calculated value, particularly in the case where CSP provides operating reserves. The assumption of a flat heat rate overestimates the performance of CSP at part load. This has little impact on the plant when providing only energy services, because the plant operates at nearly full capacity most of the time. However, when providing reserves, the plant operates at part load most of the time, and often operates at its assumed minimum. We examined the possible impact of part-load impacts by applying a polynomial heat-rate curve from Kearney and Miller (1988), where the efficiency at 40% load is about 7% lower than at full output. This curve was used to calculate the reduction in energy sales based on the marginal energy prices generated in the case with CSP. The reduction in CSP value was about $0.8/MWh.
The operational value calculated here does not include any costs associated with CSP operation. These include variable O&M, costs associated with CSP plant starts (excluding energy losses), and impacts of operating the plant at part load, including constant ramping during provision of regulation and load-following reserves.
The majority of the avoided costs is derived from reduced fuel use and associated emissions. The PLEXOS model tracks the total fuel used by all generators in the entire Western Interconnection. As discussed previously, the PLEXOS model generated by the CAISO includes the entire Western Interconnection, and there is substantial interaction between California and other Western states. The avoided fuel per unit of added generation can be tracked in the same manner as avoided costs. In each case, the total annual fuel offtake can be summed and compared to the case without the added generator. This difference is then divided by the total annual generation to produce the values in Table 4.
Of obvious note is the fact that a generator located in California can avoid a substantial amount of coal generation despite there being no significant coal-fired generators within California. This is due to a combination of factors including the inherently interconnected nature of the Western Interconnection, significant imports of electricity into California, and the modeling assumption of least-cost dispatch throughout the West.
The relationship between imports and avoided fuel is driven by the patterns of generation, load, and CSP dispatchability. From roughly fall to spring, a significant fraction of the flat block and PV generation occurs during periods of moderate load and lower imports into the state of California (compounded by wind generation and large amounts of hydro generation in the spring). As a result, a large fraction of the flat block and PV generation avoids imports, as opposed to in-state gas-fired generation. These imports in the CAISO PLEXOS model are derived from a least-cost mix of generators, which often includes a substantial fraction of coalgenerated electricity. During the first and fourth quarters, more than half of the fuel avoided by PV is coal.24 Overall, this reduces the avoided-fuel value of PV, but increases its avoided emissions value, as observed by the higher value compared to CSP in Table 3.
In contrast, the dispatchability of CSP allows it to generate during periods of highest net demand, and it tends to avoid a higher fraction ofin-state gas-fired generation. The dispatchability of CSP also increases the overall efficiency of system dispatch by providing rapid ramp rates and reserves.By smoothing the net load in California and reducing the number of partially loaded gas plants on line to provide ramping and reserves, it can increase the output of baseload units, including out-of-state coal-fired generators. Additional analysis is required to evaluate the potential limits on the market transactions effectively simulated in these scenarios. From a technical standpoint, the actual flexibility of the coal fleet in the Western states may restrict some of the operation assumed here.
The value calculated by a production cost model only addresses the variable operational value. Both CSP and PV have the ability to provide system capacity and replace new generation. However, the actual capacity value of solar technologies depends on their coincidence with demand patterns and how this coincidence changes as a function of penetration. A previous analysis of CSP plants with 6 hours of storage in California (nearly the same configuration evaluated here) found essentially 100% capacity credit using several years of data in historical systems (Madaeni et al. 2011). Capacity credit for PV generators varies depending on the year evaluated and module orientation (including the use of tracking technology), and it falls significantly as a function of penetration (Madaeni et al. 2012, Mills and Wiser 2012).
To estimate the capacity value of CSP in the 33% scenario evaluated in this report, we examined the performance of the generators during the periods of highest price, where price is used as a proxy for highest risk. Because we use only a single year of meteorology and load data, the results presented here are not generalized results; but they do provide at least some indication of the value of different generators types to provide reliable capacity. We use the capacity factor approximation, where the capacity value is approximated by the plant’s capacity factor during a set of “risky” hours. A variety of analyses have evaluated the capacity factor approximation technique to determine the number of hours that can be used to approximate more complex reliability-based approaches (Madaeni et al. 2012). These analyses have evaluated from the top
10 hours to the top 10% of hours (876), with one study suggesting the top 10 hours is closest to more robust techniques(Madaeni et al. 2012). Figure 9 shows the average CSP capacity factor as a function of the number of hours considered using the results from the PLEXOS dispatch. For CSP with thermal storage, the number of hours considered appears to be largely irrelevant in the year evaluated. CSP plants were dispatched by PLEXOS to meet demand with essentially 100% capacity value during all high-priced hours. For PV, the capacity value is about 47% using the top 10 hours and about 40% using the top 1% of hours, using the AC rating of the PV system.
Table 5 summarizes the capacity value estimates from this analysis. The first row in Table 5 is the capacity credit in terms of fraction of rated capacity. This value assumes an equal outage rate for maintenance across technologies. The second row translates this into an annualized value per installed kilowatt of the corresponding technology by multiplying the capacity credit by the low and high estimated annual value of a reference generator with 100% availability. There is a large range in estimates for the value of new capacity, with an extensive discussion provided by Pfeifenberger et al. (2012). We use a low value of $55/kW and a high value of $212/kW.
Row 3 of Table 5 translates this value per installed kilowatt into a value per unit of generation. This is calculated by multiplying the value per unit of capacity by the total capacity credit (to get the total annual value of the installed generator), then dividing this value by the total energy production. This introduces a somewhat counterintuitive outcome, resulting largely from the impact of SM and the use of TES, as demonstrated previously by Mills and Wiser (2012). The PV plant has about twice the installed capacity as the CSP plant to provide equal amounts of energy, and about half the capacity value per unit ofinstalled AC capacity; therefore, their net capacity value (as measured by unit of energy production) is similar.
The total value of the different generation sources is the sum of the operational value and
capacity value. Figure 10 summarizes the values for the different cases by combining the
operational value from Table 4 and the capacity value from Table 5.
The overall value of CSP in this analysis ranges from about $80/MWh to about $135/MWh. The range is driven by assumptions about the ability of CSP to provide operating reserves and the cost of alternative generation capacity. The ability to provide reserves added about $17/MWh, assuming that CSP plants have rapid ramp rates while operating at part load. The cost of new capacity (which may include consideration of the actual flexibility provided by new capacity) provides the largest range, with the high-capacity cost case adding about $39/MWh of value compared to the low-capacity cost case.
This variation in total value for a CSP plant also produces a large range in the value difference between CSP and the other generator types considered.Compared to a baseload plant, this difference ranges from $30/MWh to $51/MWh, whereasthe difference between CSP and PV ranges from $32/MWh to $40/MWh…
CSP with TES creates a dispatchable source of renewable energy. However, this dispatchability is constrained by the hourly flow of solar energy. As a result, modeling its value is challenging and requires chronological simulation to assess its value in providing energy, ancillary services, and firm capacity.
In this preliminary analysis, CSP was incorporated into the CAISO’s environmentally
constrained 33% RPS case and its value compared to a baseload resource and also to PV. The energy-shifting value of CSP with TES was about $6/MWh higher than a baseload resource and about $12/MWh greater than the PV resource. The difference relative to PV is influenced by the coincidence of solar supply with demand, which will change as a function of penetration and also potentially to the operational restrictions resulting from the high SM assumed in this analysis. A lower SM may be more optimal in the scenario evaluated, but the relative value of CSP and optimal CSP configuration will likely vary with the increase of renewable penetration and the decrease in coincidence ofsolar energy supply with net demand.
When CSP is allowed to provide operating reserves, its operational value increased by about $17/MWh (producing a total difference of $22/MWh compared to the baseload resource and $29/MWh compared to the PV generator). The ability to provide reserves appears to have a significant value, but will require a different operational approach for CSP—greater operation at part load and more frequent plant cycling. The additional costs of this operation, which were not evaluated here, could reduce the net benefits of CSP providing operating reserves.
Finally, in the single year analyzed, the capacity value of CSP with TES is expected to be very high, because an appropriately scheduled CSP plant would have energy available during essentially all the highest-priced demand hours of the year. The additional value provided by CSP dispatchability will depend largely on the assumed cost of alternative capacity.
Combined, the operational and capacity value of CSP calculated in this analysis ranges from about $80/MWh to about $135/MWh. This represents an incremental value of $13/MWh to $51/MWh compared to a baseload resource, or $15/MWh to $40/MWh compared to PV.
Additional analysis is needed to provide additional validation as well as explore the sensitivity of these results to additional technologies and scenarios. The relative value of dispatchable resources such as CSP with TES would likely increase as a function of VG penetration. A key element of future analysis will include exploring alternative CSP technologies and higher renewable penetration scenarios.