AI-aided optimisation and technoeconomic analysis of peaker particle-based concentrated solar power

Commercial concentrating solar power (CSP) systems depend on the cost-effective use of storage to provide a valuable service to the electricity grid. However, the tailoring of optimised ‘peaker’ systems, within the context of power purchase agreements (PPA) with variable time-of-day (TOD) pricing has received relatively limited attention. In this study, a system-level model of a particle-based CSP systems with nominal power output of 100 MW˙e is developed with detailed component-level models, a moving-window dispatch optimiser based on linear programming, and AI-based surrogate models of the receiver and power block components to accelerate calculations. The system is optimised for a range of design variables including those for field and tower layout, storage capacity and insulation thickness, for a specified TOD price schedule. System-level optimisation minimises the PPA bid price ($L_{\mathrm{bid}}$), while the dispatch optimiser maximises the TOD-weighted energy output ($\bar{T}E$). The optimal-dispatch system has a capital cost 32% lower than a system designed for immediate dispatch and minimised levelised cost of energy (LCOE), and dispatches 39% less annual electricity, but achieves an average electricity selling price that is nearly double that of the naive LCOE-optimised system. Although these results are specific to the TOD case considered here, this study highlights an integrated approach to CSP system design for high value in a realistic grid context.