Genetic Algorithm-Optimized Neural Network for Maximum Efficiency Tracking of Electrolytic-Capacitor-Less WPT System

The use of a matrix converter (MC) and a semi active rectifier in the wireless power transfer (WPT) system enables wide-range buck–boost operation and power regulation while completely eliminating electrolytic capacitors. In this paper, a dual-side closed loop control strategy is proposed to achieve high-quality grid current and constant-current/constant-voltage charging. The system is under-constrained, with the primary MC current serving as a free variable. An efficiency model is established to identify the optimal current that maximizes system efficiency under varying grid voltage, battery voltage, charging current, and mutual inductance. To adaptively track this optimal point, a genetic algorithm-optimized neural network (GANN) based maximum efficiency tracking (MET) method is developed. The genetic algorithm minimizes redundant neurons, achieving higher prediction accuracy and lower computational complexity, thereby facilitating real-time DSP implementation. Experimental results confirm that the proposed GANN-based MET method achieves high efficiency over a wide operating range.