| High-frequency resonant converters have become a recent research hotspot for soft switching and high power density.However,the increasing number of resonant components adds difficulties to resonant parameter design.The traditional parameter design process mainly relies on simulation software and human experience,and it is challenging to optimize the circuit performance fully.Therefore,this paper proposes an optimized parameter design method for high-frequency resonant converters to improve efficiency and power density,which combines Back Propagation(BP)neural networks with multi-objective genetic algorithms(MOGA).Firstly,the equivalent circuit of the isolated flyback SEPIC resonant converter is introduced,and the calculation formulas of the circuit characteristic for each operation mode are derived.The characteristic parameter curves are obtained by programming calculation,which provides the analysis basis for the circuit characteristics.Then the design method of the isolated flyback SEPIC resonant converter based on parameter sweeping is introduced,and a set of parameters is designed.It is worth noting that this method relies on human experience to optimize the performance of multiple circuits,which results in the difficulty of finding the optimal parameter design.Secondly,a multi-objective optimization parameter design method is proposed to solve the problems mentioned above.Combining the predictive model based on the characteristic parameters with the multi-objective genetic algorithm,the multi-objective genetic algorithm based on the BP neural network prediction model is introduced.Taking the isolated flyback SEPIC resonant converter as an example,the multi-objective optimization design of the converter parameters under different switching frequencies is given in this section.Thirdly,based on the parameters designed by the multi-objective optimization method,the magnetic component design and component selection of the converter at 1MHz resonant frequency are introduced in detail.The power loss of the circuit with different parameters is analyzed as well.The design methods of the planar transformer and hardware circuit under10 MHz switching frequency are introduced,as well as the selection results of the main components.Finally,an experimental prototype with 48 V input,12V/5A output,and 1MHz resonant frequency was built in the laboratory.With different design parameters,the converter has been experimentally verified.In addition,a prototype with 18 V input,5V/2A output,and 10 MHz switching frequency was built,and the experiment was carried out at a high switching frequency.Compared with the parameter sweeping method,the experimental results verify that the multi-objective optimization parameter design method shows advantages in efficiency and power density. |