TODAY’S STUDY: WHAT STORAGE MEANS TO SOLAR POWER TOWERS, SOLAR TROUGH PLANTS, AND CONVENTIONAL POWER
Estimating the Performance and Economic Value of Multiple Concentrating Solar Power Technologies in a Production Cost Model
Jennie Jorgenson, Paul Denholm, Mark Mehos, and Craig Turchi, December 2013 (National Renewable Energy Laboratory)
Dispatchable power plants provide multiple services to the electricity grid, including the ability to respond to changes in supply or demand. Concentrating solar power (CSP) with thermal energy storage (TES) is a unique source of renewable energy in that the solar thermal energy can be dispatched similarly to conventional thermal generation. However, CSP-TES plants are an energy-limited resource, meaning that their response might be restricted by solar availability. Therefore, the use of this solar energy must be optimally scheduled to provide the greatest value to the system. The timing of CSP-TES dispatch depends on a variety of factors, including electricity demand patterns, the penetration of variable generation (VG) sources, and the configuration of the CSP plant itself.
Recent studies have identified the value of CSP-TES in not only shifting energy over time but also in providing ancillary services and serving as a source of firm capacity (Denholm and Hummon 2012; Denholm et al. 2013; Denholm et al. 2012). However, previous analysis simplifies several operational parameters of CSP as a thermal power plant and does not consider multiple CSP technologies and configurations now being deployed.
We use an established CSP modeling framework in a commercially available production cost model to compare the operation and value of two CSP technologies: molten salt towers and parabolic troughs. In addition, we consider a range of configuration parameters, such as solar multiple (SM) and storage size, to evaluate how the operational and capacity value varies with plant configuration.
Impact of Capacity on the Overall Value of CSP-TES
The calculated value for the various configurations of tower CSP-TES presented up to this point considers only operational value and does not consider the ability of CSP-TES to displace new conventional thermal generation. The actual capacity value for a VG resource depends on the coincidence of resource availability with net load. For CSP-TES we use an approach from Tuohy and O’Malley (2011) for estimating capacity value from devices with storage. This approximation considers the stored energy available for dispatch during hours with the highest probability of unserved load. For instance, if the CSP-TES plant is dispatched below capacity but has enough stored energy to generate at full output during that hour, the device earns a 100% capacity credit during that hour.23 In this simulation, we consider the 88 hours (or 1%) with the highest net load (equal to load minus the contribution of wind and solar PV). The capacity credit, determined with this “maximum generation” approach, is 100% for all configurations with storage. Additional discussion of the capacity value of CSP plants is provided by Madaeni et al. (2012).
The corresponding monetary value of this capacity can be based on the cost of comparative conventional plant. We use a range of values with a low annualized cost of new capacity equal to $77/kW-yr and a high value of $144/kW-yr (Xcel 2011). This capacity value is contingent on a system actually needing additional capacity to provide an adequate planning reserve margin—for example, to replace a retiring generator or to meet growth in demand. A system with an adequate planning reserve margin would have essentially zero capacity value for a new resource.
Table 10 shows the capacity value for each CSP-TES plant for the year of operation. The first column shows the rated capacity of the plant, assuming equal energy production. Because all plants with storage are assumed to have a 100% capacity credit, their capacity value is simply proportional to their size. This produces the result that the lower SM will inherently be worth more as a system resource because lower SM plants have larger megawatt capacities in this analysis. The size of the plant, multiplied by the assumed cost of new capacity (either $77/kW-yr or $144/kW-yr) produces an annualized capacity value in the third results column. Finally, this value is divided by the annual energy production to derive a capacity value per unit of energy ($/MWh) in the last column. The annualized capacity value will be the same for all plants with the same SM regardless of storage size. However, the capacity value per unit of energy will actually decrease for the plants with more hours of storage due to their slightly greater energy production resulting from lower spillage. This demonstrates another limitation of using the value per unit energy performance metric.
The full value of each plant is the sum of its operational value and its ability to offset additional new capacity (capacity value). Figure 14 adds the operational values in Figure 13 with the capacity values measured in dollars per megawatt-hour from Table 10. Because the operational values in Figure 13 are relatively flat for each SM, we only use the “base” configurations described in Table 10, which includes the SM = 1.3 with 3 hours of storage case. It includes four values for each configuration—combinations of low and high operational value using the base and double NG price case (2x NG), and the low and high capacity value from Table 10. Figure 13 and Figure 14 indicate that smaller SMs may provide more system value.
The ability to vary both SM and hours of storage is an important aspect of CSP plant design. Adding energy storage enables reduction in LCOE due largely to decreased spillage. However, the optimum CSP design should consider not only cost but also the value of energy and capacity CSP delivers to utilities and system operators. This analysis demonstrates that multiple CSP technologies (both troughs and towers), as well as plant configurations, can be analyzed using traditional planning tools such as production cost models.
We find that a parabolic trough CSP-TES plant may require a higher capacity, at a greater expense, than a similar molten salt power tower to achieve the same annual output largely due to a larger seasonal variation in output, lower thermal efficiency, and greater storage losses. However, we find that the system value as measured by their value per megawatt-hour of delivered energy of dry-cooled tower and parabolic trough CSP-TES plants are similar despite their different solar resource profiles.
We also analyzed various configurations of SM and thermal storage capacity. We found that lower SMs (with correspondingly larger rated plant capacity) had the largest marginal system value and benefitted from some additional storage to prevent spilled energy. However, for all SMs, we found only a small benefit to storage from 6 to 9 hours of rated plant capacity and less benefit beyond 9 hours. Plants with smaller SMs shifted energy more effectively to periods of high load due to their increased capacity, acting as “peaking” units, while plants with higher SMs acted more like base-load plants, generating at a constant output for more hours of the day.
Plants with smaller SMs incur additional capital costs for a larger power block but may also be able to earn more capacity value. Capacity is likely an important source of value for CSP plants, which appear to offer a capacity value similar to conventional thermal plants if properly scheduled and utilize accurate forecasts of solar availability.
Further analysis will examine the role of energy-limited resources such as CSP-TES in providing ancillary services under current and alternative sets of regulation and markets. The relative value in CSP-TES, and the capacity factor of the CSP-TES plant, is dependent on renewable energy penetration (including existing CSP-TES plants). For instance, this system has a relatively low penetration of PV generation, which is largely coincident with CSP solar resource availability. In addition, further analysis will determine the system needs at shorter timescales as VG integration continues to influence sub-hourly system operation.