Study on multiple resonant points control strategy of contactless power transfer system using neural network

This paper presents a power control strategy based on multi-resonant operating points, which is realized by multilayer feedforward neural network applying back-propagation algorithm. The full-bridge contactless power transfer system and magnetizing current of the transmitter on primary side are respectively used as research object and control variable. After batch-learning and training, the converged network determines alternating operating duty cycle of each resonant operating point in one cycle. By controlling magnetizing current, it can fulfill the dynamic regulation of transmission power, and increase the energy transfer efficiency. Simulation results show that in the control strategy, magnetizing current on the primary side can be stabilized at any set value for given range, and system has some disturbance restraint performance, satisfied with contactless power transfer system control demands.