A model of the solar field for the chosen plant was developed by tuning previously created models of trough plants by the research team. Most of the data required for tuning was provided by the involved company; however, some specific tests were necessary to fine-tune valve behavior.
The market price strategy control algorithms, developed in the advanced grant, had to be tuned and simplified for implementation on the solar plant's Distributed Control System (DCS) with the available information. The total heat transfer fluid (HTF) flow in the solar field was regulated by an MPC controller, which controlled the overall solar field temperature. The controller distributed the HTF flow across sectors and loops to maximize the collected energy while minimizing defocusing actions, thereby reducing maintenance costs. Artificial neural networks (ANNs) were used to approximate the controller. The resulting approximated controller was tested through extensive simulations using the dynamic model of the plant's solar field.
The controller was then programmed into the solar plant’s DCS, and tests were performed as planned. Figure 1 presents the results of an experiment conducted in the West 2 sector of a 50 MW plant (HelioEnergy 1). As shown, after the controller was activated at 12:15, the hottest loop "bought" oil allowance from the cooler loops by opening their loop valves, while the least efficient loops "sold" their HTF allowance to the most efficient loops by closing their loop valves. This process allowed the most efficient loops to receive more oil than the less efficient ones, thus maximizing the amount of collected energy. This gain is also evident from the increase in the solar radiation collection factor of the solar field, which rose by approximately 2%.
