Rotor Stress Limitation in CSP Plants Through Reinforcement Learning Techniques

Concentrated solar power plants are actually playing a crucial role in the energy transition. The gradual phase-out of fossil fuels makes them a promising technology to include in a low-emission energy mix. The increasing use of these plants has highlighted their current critical issues and the changes that need to be made in control systems to solve problems that are negligible in conventional thermoelectric plants. One of the most critical components is the turbine rotor, which is subject to high thermal gradients several times a day, due to the variability with which the sun heats and generates steam. In the present work, a control architecture, that combines power regulation and rotor stress limitation, is developed in the Simulink environment of Matlab2022b. The controller design is based on the dynamic response of the thermomechanical model of the rotor. A Reinforcement Learning approach is used to optimize the controller parameters with the PGPE algorithm.