Machine Learning Assisted Parameter Tuning for a L2CL-LCL Compensation WPT System

摘要

Nowadays, wireless power transfer (WPT) has received a lot of attention due to its inherent advantages, such as convenience, safety, low maintenance, weather proof, etc. However, the parameter tuning is critical in conventional techniques of L2CL - LCL compensation WPT systems as they have difficulty in system designing and poor capability of higher order harmonic suppression. In order to overcome these deficiencies, a novel machine learning NSGA-II (Non-dominated Sorting Genetic Algorithm) based optimization topology has been proposed for wireless power transfer. This is a derivative free open-source circuit optimizer which designs circuits simpler than ever before. It optimizes the drawing current and system efficiency as specified by the user. Implementation was carried out using the Python Language under the simulator of "Python Power Electronics". To verify the efficiency and feasibility of the proposed NSGA-II optimization method, it is compared with the conventional tuning of the L2CL-LCL Compensation topology. The overall efficiency of the proposed system has been increased from 95.2% to 98.5% compared to the conventional method.