Research On Current Tracking Control Method Of Single-phase Grid-connected Inverter Based On Reinforcement Q-learning

During the operation of the single-phase grid-connected inverter,due to disturbances from the DC side and AC side and the presence of parameter uncertainty,the control performance of the inverter will decrease.In actual engineering applications,the parameter information may not be completely available,and even in some cases may deviate from their rated values due to aging and other reasons.In these cases,the controller designed based on the system parame-ters cannot be used.Therefore,this paper introduces the idea of reinforcement learning in the control of single-phase grid-connected inverters,and uses the reinforcement Q-learning algo-rithm to design a controller that does not depend on system parameters,and ensure good control performance.The research work of this article mainly includes:1)For the single-phase grid-connected inverter,in order to improve the current tracking control effect,a resonance controller is added to eliminate the tracking steady-state error of the inverter system output current.The H_∞tracking control method is selected to design the controller.By designing the discounted performance function,establishing the discounted game algebra Riccati equation,and solving the game algebra Riccati equation,the H_∞tracking control law can be obtained to achieve a good control effect.2)Design an inverter controller based on state feedback Q learning algorithm.The con-troller is designed with a Q learning algorithm.It is no longer necessary to obtain circuit param-eter information.Instead,iterative learning is performed by collecting system state information to solve the recursive Bellman equation.The final iteration will be approximately converge to the solution of the game algebraic Riccati equation,so as to realize the inverter current tracking control based on the state feedback Q learning algorithm.According to the equivalence of the Q function and the discount performance function in the H_∞tracking control,it is proved that the Q learning control algorithm can achieve the same control effect as the H_∞tracking control method.3)Design an inverter controller based on output feedback Q-learning algorithm.The state reconstruction method is adopted to re-establish the Q function using input and output data.The input and output historical data of the system is collected to reconstruct the state information of the system,and obtain the optimal control law through iterative solution.Compared with the state feedback Q learning method,the output feedback Q learning method does not require obtaining all the state information of the system.In the actual grid-connected inverter system,the full state measurability is a more difficult requirement to achieve,and secondly.The output feedback Q learning method can reduce the sensor Number,thereby reducing the complexity of the system and improving the reliability of the system.