An Integrated Computing Method For Electrical Equipment Based On Cloud Computing

The design and optimization of modern electrical equipment need to consider design factors such as fine simulation,manufacturing process constraints,service characteristics and new material characteristics,making it a computationally difficult,multi-temporal and nonlinear problem that needs to be solved.Traditional methods generally use simplified finite element model to reduce the need for computing resources to achieve performance analysis and rely on human experience to implement the necessary interventions in the optimization process to achieve performance optimization,resulting in unsatisfactory aspects such as calculation time,result accuracy and practicability.On the one hand,this thesis uses the elastic,dynamic and easy to expand cloud platform to solve the large-scale electromagnetic finite element and multi sample task calculation problems in the optimization design of electrical equipment,so as to reduce the calculation time;On the other hand,considering that the support vector machine has superior data feature expression ability and strong generalization ability when dealing with small sample data and nonlinear problems,the support vector machine is used to perform nonlinear regression analysis on the electromagnetic model,then use intelligent algorithms optimize its structure to improve calculation accuracy and credibility.Therefore,this thesis proposes an integrated computing method for electrical equipment based on cloud computing,the main research contents are as follows:(1)The basic theory of multi-objective optimization problem,the principle and characteristics of genetic algorithm,non dominated sorting genetic algorithm(NSGA)and non dominated sorting genetic algorithm with elitist strategy(NSGA-?)are analyzed and studied,and the NSGA algorithm and NSGA-? algorithm are compared and tested by using the standard function with theoretical reference value and multi peak value.(2)The cloud platform finite element test environment is built,three finite element calculation cases such as TEAM Problem 7,three-dimensional inductance and three-dimensional motor are run,and the CPU core number and memory configuration with the optimal calculation time and resource consumption are obtained.The three objectives of the multi-objective task scheduling problem of electromagnetic cloud computing,namely,the maximum calculation completion time,the total machine load and the maximum machine load,are optimized by NSGA-? algorithm,The most satisfactory task scheduling scheme of cloud platform is selected from Pareto solution set by multi index weighted grey target decision model.(3)Starting from the finite element simulation calculation of transformer,through the sample space design of transformer parameters,the nonlinear regression analysis of its electromagnetic model is carried out by using support vector machine,and the NSGA-? algorithm is introduced to optimize its agent model to obtain the most reasonable set of transformer structural parameters,and the corresponding finite element model is established for simulation and verification.The results show that cloud computing platform can provide a lot of computing resources for multi sample tasks.The support vector machine and NSGA-? algorithm were used to optimize the structural parameters of the transformer.The established proxy model was accurate,feasible and efficient.The reliability of the optimization results was verified by the finite element model.It is proved that the support vector machine and NSGA-? algorithm are correct and practical for the optimization of transformer structural parameters.