Compatibility and Performance Improvement of the WPT Systems Based on Q-Learning Algorithm

Different kinds of compensation topologies are widely used in wireless power transfer (WPT) systems, resulting in loads with different compensation topologies hardly obtaining the same power from the same power transmitter. In addition, the load power and system efficiency are hard to hold constant when the mutual inductance or load resistance varies, and these two variation parameters are both hard to be accurately identified. These severely limit the practical popularization of the WPT technology. Therefore, a new strategy to improve the compatibility and performance of the WPT system based on the Q-learning algorithm and switch-controlled-capacitor (SCC) is proposed in this work. The Q-learning algorithm is used to train an intelligent offline database by monitoring the voltage and current in the primary side of the system to obtain an optimized compensation capacitor on the primary side without knowing the mutual inductance and load resistance. A 1-kW prototype is established and the experimental results demonstrate that the load side can receive desired power when the load changes from one kind of compensation topology to another, or the load resistance and mutual inductance vary together. Meanwhile, the coil-to-coil efficiency can also be effectively maintained.