Constant Voltage and Constant Current Control Strategy for WPT Systems Based on Dynamic Programming Reinforcement Learning Algorithm

Different kinds of compensation topologies in wireless power transfer (WPT) systems usually can only achieve constant voltage (CV) or constant current (CC) output at the same time when the load condition varies, such as during the charging process of a battery. This seriously impacts the stability and application of the WPT systems. Therefore, dynamic programming reinforcement learning (DPRL) algorithm is proposed in this work to achieve CV and CC outputs in the common LCC-LCC/S WPT systems. The proposed method does not need parameter identification or accurate mathematical model, and can thereby reduce the complexity and improve the accuracy of the control. An intelligent offline database is trained based on DRPL algorithm by only monitoring the voltage and current in the primary side of the system to obtain an optimal compensation capacitor on the primary side. A kilowatts experimental setup is built, and the results show that the 180 V CV or 5.42 A CC output can both be realized as expected in each system according to the requirements of the load conditions. The maximum and average errors are only 1.29% and 0.93%, respectively, with nearly no impact on the system efficiency, which effectively validates the feasibility of the proposed method.