Computational Optimization of Solar Thermal Energy Generation and Storage

Integrating renewable energy resources into the power generation system is essential for achieving future renewable portfolio standards. Solar thermal is not as popular as solar photovoltaic, but it does enable thermal energy storage (TES), which can provide longer durations of storage than many other technologies. The expense of installation is a significant drawback of solar thermal with TES, but the cost is declining over time. Permitting these operators to sell power at retail utility rates may improve the economics of solar thermal projects.

In this work, we model and optimize a 15 MWe parabolic trough concentrating solar power plant with thermal energy storage. The model simulates power output of a 10 hectare solar field assumed to operate in Daggett, California. We use a proxy model for the solar field, and simulate a three-stage heat exchange process, a thermal energy storage tank with radiative, convective, and conductive heat loss, and a steam turbine. Irradiation data from the National Solar Radiation Database (NSRDB) [4] and electricity price data from the California Independent System Operator (CAISO) [1] are clustered into representative days for the year 2017. We use the Particle Swarm Optimization (PSO) - Mesh Adaptive Direct Search (MADS) hybrid optimization scheme to solve for the optimal operations of the plant. Decision variables include the mass flow rates into and out of storage at hours 7-22 for each clustered day. Constraints include the pinch temperature, approach temperature, steam quality, and temperature ranges of the heat transfer fluid (HTF).

We first consider idealized price curves, for which the optimal operations are intuitive, and observe that optimizer performance agrees with intuition for these cases. We then use the CAISO price data and evaluate the value of two-hour, four-hour, and six-hour storage tanks by optimizing operations over each clustered day to compute the annual profit and net present value (NPV) of a plant over 30 years. We find that the four-hour storage tank is optimal, while the high installation cost of the six-hour storage tank renders it suboptimal despite a higher annual operating profit. Lastly, we conduct a sensitivity analysis to compare the base case, a “fatter” duck price curve, and Power Purchase Agreements (PPAs). We find that a six-hour storage tank yields the highest NPV with the more volatile prices of the “fatter” duck curve, as the higher peaks in electricity prices offset the higher cost of a six-hour storage tank. A plant operating under a PPA without storage is found to be the least profitable, suggesting that it is beneficial to utilize TES and sell at market prices rather than enter into a PPA.