Study On Power Control Based On Neural Network Of CPT System

Contactless Power Transfer technology is a new type of electric power transmission technology that utilizes power electronic technology, magnetic coupling technique and control theory, and obtain electricity from grid access by the way of no-electrical touch. It is safe, convenient, easy to maintain, and also has high reliability and strong environmental affinity, etc. Due to the complex nonlinear system characteristic of CPT systems, it is difficult for traditional control method to meet control requirements. With the development of the intelligent control theory, the artificial neural network has increasingly attracted more attention. Especially the arbitrary nonlinear systems approximation characteristics of the neural network, make neural network a new subject in control theory.The original edge LCL and the vice edge series resonance type and the original and vice edge series resonance type of CPT systems were studied. First of all, the system was analyzed by making the modeling of cycle strobe mapping and periodic fixed point analysis. The operating point of soft switch was determined by periodic fixed point. More detail about characteristic analysis and simulation validation were performed.Secondly, the neural network power control strategy was proposed based on power transmission characteristics of the soft switch set-point. BP neural network controller makes CPT system automatically switch back and forth among the soft switch set-points, and controls the power more effectively. The output voltage and current of original edge were studied in this article, the appropriate soft switch set-point were chosen by selective work cells of frequency size, which regulates control signals of input pulse width modulators, then adjust the switch time length of switch tube and adjust output voltage of power converters in the end.Since the BP network have the defects of long convergence time and easily into the local minimum value, we use Particle swarm optimization algorithm to optimize the weights of BP network and improve the performance of BP network controller, to a certain extent, which speeds up the process of the BP algorithm convergence, and then improve the BP neural network training and inspection accuracy. Finally, the designed neural network controller was simulated and validated, the simulation results show that through the control of the original edge resonant current, RMS of the original edge resonant current can be stabilized on any value within a given range, and exhibit certain resistance to disturbance; and through the output voltage control, the output voltage can be still stable on a given value with the changing load. The system has strong robustness, meet the control requirements of CPT system and soft switch conditions, achieving higher transmission efficiency.