Efficiency Optimization Of High Frequency Cascade Converter Based On Genetic Algorithm

In recent years,the cascade converter has been widely used in electric vehicles and solar power system due to its strong voltage regulation capability.However,the complex control parameters make the traditional efficiency optimization method difficult to apply.Genetic algorithm(GA),as a black box optimization algorithm,can optimize targets when the circuit model is unknown,which brings new options for this problem,but there are still two problems to be solved: First,how to make the GA control loop work in coordination with other control loops to avoid system offset;Second,how to accelerate the convergence speed of the algorithm and avoid it falling into the local optimal solution to meet the high frequency demand.In this thesis,the IBBLLC cascade converter is optimized.For the first problem: Firstly,due to the fact that the IBBLLC converter is easily detuned with the change of phase shift,and the deadtime is related to the ZVS of the latter LLC stage,loss and working status are analyzed and then the boundarys of the two parameters are calculated,which prevents the system detuned during the GA optimization.Secondly,in order to coordinate the PID control loop and the GA optimization loop,timers of MCU are set,which enables accurate efficiency can feedback.For the second problem: This thesis adds an optimized elitism selection strategy to GA,which ensures that the individuals with the highest efficiency are always retained,thus local convergence is avoided and the convergence speed is also improved.Based on the MATLAB-simulink simulation platform,the feasibility of the above solution is verified.Under the condition of 425V/380 W,The convergent efficiency of optimized GA is 2% higher than the traditional one,which alleviates the local convergence problem.On the basis of veification of simulation,the optimization strategy is verified in the 1440W/1MHzIBBLLC prototype.The results show that under the input voltage of 260 V,the average efficiency is improved by 1.15% and efficiency achieve convergence within 200 generations whith the optimized GA.