Process integration and optimisation of particle based concentrated solar power system

Abstract

This thesis presents physical and technoeconomic modelling studies exploring optimised configurations of particle-based CSP in various scenarios. Detailed component models were developed and integrated into a dynamic system-level model using SolarTherm, validated and referenced extensively. An optimisation tool based on genetic algorithms (GA) was developed to ensure global optima in the optimisation. The study also explored an artificial intelligence (AI) surrogate modelling framework for several CSP components, significantly reducing computational expenses by nearly 100X with minimum error and increasing the numerical stability of the simulations. Three analyses are presented in this thesis. The first is an analysis of the optimum configuration of a particle-based CSP to produce a constant electricity supply (base-load plant), with three different receiver arrangements: single aperture (SA), parallel aperture (PA), and cascaded aperture (CA). It is found that both CA and PA systems yield the lowest LCOE, 57.96 USD/MWhe and 57.97 USD/MWhe respectively. The PA system has 12.83 hours of thermal energy storage (TES), while the CA system has a slightly bigger TES, 13.75 hours. Of all these three proposed systems, CA has the highest CF of 79.57%, while PA and SA have 79.15% and 77.99% CF. The second analysis is an optimisation of a single aperture particle-based CSP system operated as a peaker plant in a specific time-of-day (TOD) electricity market. A dispatch optimisation based on linear programming (LP) was implemented to ensure the dispatched electricity yields the maximum possible revenue over a certain forecast horizon. A side-by-side comparison of the optimum SA peaker plant with the optimum SA base-load plant shows that the optimum peaker plant yields 37% less electricity for an identical power block size. However, the peaker plant accumulated almost 20% more revenue than the base-load plant. It is also shown that the optimum peaker plant has a much smaller system: a shorter tower (270.2 m for base-load, 184.1 m for peaker), smaller TES size (16.05 hours for base-load, 14.68 hours for peaker), and a smaller heliostat field (3.71 solar multiple for base-load, 2.05 solar multiple for peaker), which in total contributes to a 157 MUSD cheaper peaker system. The third analysis explores the role of particle-based CSP in the industrial decarbonisation scenario. The analysis was conducted in two steps: a hybrid plant to produce cold H2, followed by co generation of heat and H2, Even though the cheapest cold H2 was produced by the PV-only plant (no CSP) at 2.98 USD/kg-H2 , between 40%-80% CF, the optimisation prefers to increase the size of the CSP system while holding the size of the PV system almost constant to increase cold H2 production, signifying that particle-based CSP has a role to play at higher CF. Even though the cheapest cold H2 was produced at CF 35%, an analysis of production loss costs revealed that the optimum hybrid plant was the configuration that yields CF 88% 420 MW th CSP system, 280 MW th PV system, and 3250 MWth h TES, for a fixed 100 MW e AEL name plate size. In the hot H 2 analysis, three different dispatch strategies were derived using LP approach due to the fact that the TES assumes two roles: as an energy storage for the power block to produce electricity and heat storage to preheat the cold H 2 to the desired temperature. The result of the optimisation shows that the best configuration which produces the cheapest reliable hot H2 is unfortunately a PV-only system (no CSP) consisting of 2.3 GWe PV, 1.6 GWe AEL, 129 hours of TES, and 145 hours of H2 storage, with no power block at LCOH2 3.1 USD/kg. Extra analyses on different heat demand were conducted to see which heat duty particle-based CSP might come into the picture. It was found that at a heat demand of 165 MWth , hybrid PV-CSP yields cheaper LCOH2 compared to the PV-only system.