A Constant Power Loads Voltage Controller Based on the Coordination of Learning and Feedback Linearization

Feedback linearization controllers are widely used in buck converters to regulate the voltage of a power system integrated by a constant power load (CPL). However, traditional feedback linearization controllers have two drawbacks. First, feedback linearization controllers are usually designed based on an approximate model of the buck converter rather than a detailed model. Second, feedback linearization controllers are practically model-based. When the parameters in the buck converter are mismatched, the performance of the feedback linearized controller deteriorates. To deal with the problems of model approximation and parameter mismatch, a reinforcement learning compensated feedback linearization (RL-FL) controller is proposed in this paper. The proposed RL-FL controller utilizes a deep deterministic policy gradient (DDPG) to compensate the control actions obtained by the feedback linearization (FL) controller to improve the performance of voltage regulation. Subsequently, a case study is examined to verify the performance enhancement of the proposed RL-FL controller compared to the FL controller.