DDPG-based heliostats cluster control of solar tower power plant

Abstract

The control of heliostat is crucial for the development of solar tower power plant. Currently, most power plants use open-loop control, which has low cost but low efficiency, closed-loop control has high concentrating efficiency, but each heliostat requires sensors and has high cost, and Proportional-Integral-Derivative (PID) controller has good control effect, but the parameter adjustment is difficult and overshooting problem occurs. In this paper, we propose a DDPG-based heliostat cluster control aimed at improving the heliostat control effect and reducing the control cost. A leader-follower strategy is used to control the heliostat, where the whole heliostat field is divided into several groups, each group is assigned a leader heliostat, and the rest of the heliostats follow the leader heliostat to rotate. The leader acquires the control error by means of a photoelectric sensor or a camera device. The following heliographs rotate with the leader to obtain the control signal, so there is no need for sensors, which reduces the number of sensors and lowers the cost. To address the shortcomings of traditional PID, we propose a DDPG-based PID control algorithm. The algorithm is trained to find out the optimal value at each moment, which ensures that the controller parameters are optimal at each moment. The results show that the tracking error is below 0.0001 rad for both cluster control and individual control. This ensures effective tracking performance while reducing the sensor cost. The controller based on the DDPG algorithm eliminates overshoots, reduces errors, and shortens the stabilization time by 0.5 seconds.