Optimum design of permanent magnet synchronous generator based on MaxPro sampling and kriging surrogate model

Abstract

In order to improve the efficiency and quality of a permanent magnet synchronous generator (PMSG) for vehicles, the parametric model and finite-element model of the generator are established. The influence of seven structure parameters on three performance parameters is obtained by simulation analysis, and the sensitivity values of the parameters are given. According to the simulation results, it can be seen that the structure design parameters have higher degrees of freedom, and the relationship among parameters is complex. Using three multidimensional functions as test functions, the precision of the fitting method proposed in this paper is compared with that of the traditional methods. The test results show that, compared with the other methods, maximum projection (MaxPro) sampling and the Kriging method can select as fewer samples as possible while guaranteeing the higher fitting precision. Then, the Kriging surrogate model, which reflects the nonlinear relationship between the structure and performance parameters, is established, and the prediction precision of the model is verified. Based on the surrogate model, the performance of the generator with three performance parameters as optimized objects is optimized by use of elitist nondominated sorting genetic algorithm (NSGA-II), and a Pareto optimal solution set of the objective function is obtained. Three of the optimal solutions are selected for finite-element verification. The results show that the Kriging surrogate model can provide better global and local prediction between input and output variables. It is further proved that the samples obtained by MaxPro sampling can fully reflect the characteristics of the design space. The optimized objects can be weighed according to the optimal decision based on the results of multiobjective optimization through the Kriging surrogate model. The effectiveness of the optimization method is verified. It is of great engineering significance to improve the performance of generators, and the optimal model obtained by this method has an important reference value for prototype trial production.

6σ Robust Lightweight Design of the GIL/GIS Bus Based on Co‐Simulation of Insulation, Current Flow, and Mechanical Stress

“…Compared with the Monte Carlo method, the Latin hypercube method uses out‐of‐order stratified sampling, which can cover the entire sample space more comprehensively through a small number of sample points [34]. The Kriging model‐fitting method is a regression algorithm based on covariance function for spatial modeling and interpolation, which has been widely used in product structure design of many industries [35,36]. The coefficient of prognosis ( CoP ) is a model quality assessment method independent of the response surface.…”