A PV Power Grid-Connected Inverter Current Tracking Control Scheme via Data-Driven Q-Learning Algorithm

Abstract

In this paper the output current tracking control problem for a PV grid-connected inverter is considered with unknown its dynamics. The performance of the controllers designed with the rated parameters will decrease when the inverter's parameters change for some reasons as the decay of its components. As such an adaptive optimal controller, which does not depend on the prior parameters, is proposed in this paper based on the data-driven off-policy Q-learning algorithm. An augmented system composed of the inverter system and the reference current generator is first established. Then, a discounted game algebraic Riccati equation (GARE) with respect to the augmented system is given assuring the solution to the tracking control problem. Finally, the off-policy Q-learning algorithms are used to solve the discounted GARE without knowing system dynamics which in turn are employed to obtain the optimal state feedback policies. Simulation results are presented to verify the theoretical analysis and demonstrate the controlled inverter's performance.