The purpose of this analysis is to explore the potential of CSP to provide grid flexibility and enable increased solar penetration in the Southwestern United States. To perform this preliminary assessment, we use the REFlex model, which is a reduced form dispatch model designed to examine the general relationship between grid flexibility, variable solar and wind generation, and curtailment (Denholm and Hand 2011). REFlex compares hourly load and renewable resources and calculates the amount of curtailment based on the system’s flexibility, defined as the ability for generators to decrease output and accommodate variable generator sources such as solar and wind.
California is a likely candidate for large-scale deployment of both PV and CSP, and has strong solar incentive programs and a renewable portfolio standard. However, modeling California in isolation ignores the fact that California has strong transmission ties to neighboring states, including Arizona and southern Nevada, which have significant potential for solar energy. Currently, power exchanges between neighboring areas in the western United States are accomplished through bilateral contracts, and typically do not occur in real time. This analysis assumes the eventual availability of real-time power and energy exchanges across California, Arizona, New Mexico and Southern Nevada to allow sharing of solar resources. It also assumes that transmission is accessible to all generation sources on a short-term, non-firm basis. This “limiting case” allows for examination of the best technical case for solar deployment without market barriers or transmission constraints.
We began our simulations by evaluating the limits of PV, given flexibility limits of the existing grid. The simulations use solar, wind and load data for the years 2005 and 2006. Load data was derived from FERC Form 714 filings. For hourly PV production, we used the System Advisor Model
(SAM), which converts solar insolation and temperature data into hourly PV output (Gilman et al. 2008). Weather data for 2005 and 2006, was obtained from the updated National Solar Radiation Database (NSRDB) (Wilcox and Marion 2008). We assume that PV will be distributed in a mix of rooftop and central systems (both fixed and 1-axis tracking). Additional description of this mix, including geographical distribution is provided in Brinkman et al. (2011).
Because California has significant wind capacity installed and plans for more, we also consider the interaction between solar and wind generation. Simulated wind data for 2005 and 2006 for California/Southwest sites was derived from the datasets generated for the Western Wind and Solar Integration Study (WWSIS) (GE Energy 2010). We started with a base assumption that wind provides 10% of the region’s energy based on the “In-Area -10% Wind” scenario from the WWSIS. These data sets were processed through the REFlex model to establish base relationships between grid penetration of PV, curtailment, and grid flexibility. The overall system flexibility was evaluated parametrically, starting with a base assumption that the system is able to accommodate PV over a cycling range of 80% of the annual demand range. This corresponds to a “flexibility factor” of 80%, meaning the aggregated generator fleet can reduce output to 20% of the annual peak demand (Denholm and Hand 2011). This value is based on the WWSIS study and corresponds roughly to the point where all on-line thermal units have reduced output to their minimum generation levels and nuclear units would require cycling. The actual flexibility of the U. S. power system is not well defined, and this value is not intended to be definitive, but is used to represent the challenges of solar and wind integration and the possible flexibility benefits of CSP/TES.
Figure 4 illustrates the framework for this analysis, showing the simulated dispatch over a 4-day period (April 7-10). It demonstrates a case where 10% of the annual demand is met by wind and 20% is met by solar. The figure shows both the simulated solar profile and its contribution to meeting load. Because of relatively low load during this period, PV generation exceeds what can be accommodated using the assumed grid flexibility limits. This typically occurs in the late morning, before the demand increases to its maximum in the afternoon. In these four days about 16% of all PV generation is curtailed and about 5% of the annual PV generation is curtailed.
FIGURE 4: Simulated system dispatch on April 7-10 with 20% contribution from PV generation and resulting curtailment due to grid flexibility constraints
Figure 5 illustrates the average and marginal PV curtailment rates as a function of PV energy penetration for this initial scenario. It should be noted that the x-axis shows penetration of only solar PV. Because wind provides 10%, the total penetration of variable generation is 10% plus the penetration of solar. The average curve shows the total curtailment of all PV at a certain generation level. At the overall assumed system flexibility level, by the time PV is providing 22% of total demand, about 6% of all potential PV generation is curtailed.
The actual allocation of curtailment strongly influences the economics of PV and other variable generation. Figure 4 also shows the marginal curtailment rate, or the curtailment rate of the incremental unit of PV installed to meet a given level of PV penetration. If curtailment were assigned on an incremental basis at the point where PV is providing 22% of total demand, only about 50% of this additional PV would be usable, with the rest curtailed.
In this analysis we “assign” all incremental curtailment to solar, partially based on the federal production tax credit which incentivizes wind
FIGURE 5: Marginal curtailment rates of PV in a base scenario in the southwestern United States assuming an 80% system flexibility
generation, while the primary federal incentive for solar is an investment tax credit that incentivizes installations but not generation. 
Curtailment of solar may also occur if wind is installed “first” and a “last in, first curtailed” rule applies. The actual allocation of curtailment is, and is likely to continue to be, a contentious issue. Regardless of allocations rules, increased grid flexibility will be needed to minimize curtailment if solar is expected to play a “primary” role in reducing fossil-fuel use in the electric sector.
The estimation of the marginal curtailment rate is important because it helps establish the optimal mix of generators serving various portions of the load. This can be observed in Figure 6, which translates curtailment into a cost of energy multiplier. This multiplier—equal to 1/(1-curtailment rate)—can be applied to the “base” LCOE of electricity generation (no curtailment). This represents how much more would need to be charged
for electricity based on the impact of curtailment and the corresponding reduction in electricity actually provided to the grid.
Both the average and marginal multipliers are shown in Figure 6. The average multiplier is applied to all PV generators. The marginal multiplier is applied to the incremental generator, and is more important when determining the role of storage or other load-shifting technologies. For example, at the point where PV is providing 25% of the system’s energy, the curtailment of all PV (average curtailment) is about 17% and the resulting cost multiplier is 1.2. If the base cost of PV is $0.06/kWh, the overall, system-wide cost of PV would be $0.06 x 1.2 or $0.072/kWh. This overall cost may be acceptable, but the costs are greater at the margin. For example, the last unit of PV installed to reach the 25% threshold has a curtailment rate of about 68% and a cost multiplier of 3.1. At a $0.06/kWh base price, this incremental unit of PV generation would have an effective cost
of more than $0.18 per kWh. This would likely result in examining options to both increase grid flexibility (to accommodate more PV with lower curtailment rates) and improve the solar supply/demand coincidence.