Design Optimization of an Interior Permanent Magnet Synchronous Machine Applying the PSO Algorithm to a Surrogate Model Based on Artificial Neural Network

Medrano Luna I.S.; Ribeiro B.C.; Cury Alves Dos Santos S.T.; Alarcon W.M.; Boffino De Almeida Monteiro J.R. · COBEP 2023 - 17th Brazilian Power Electronics Conference and SPEC 2023 - 8th IEEE Southern Power Electronics Conference, Proceedings (2023) · DOI: 10.1109/spec56436.2023.10407848

Resumen

The development of the finite element method (FEM) made it possible to refine the design of electrical machines, due to its faithful representation of this equipment. However, it requires significant computational effort, resulting in a time consuming simulation. This work explores a machine learning approach based on a multilayer perceptron (MLP) artificial neural network (ANN) to implement a surrogate model. The objective is to replace the finite element model for an interior permanent magnet synchronous machine (IPMSM) with the one proposed here. The input and output data for training the ANN are acquired using the open-source software Pyleecan. Subsequently, Particle Swarm Optimization (PSO) is applied to optimize the machine design by adjusting its constructive parameters iteratively. The ascertained mean electromagnetic torque was improved by 1.53%, whilst the sum of magnetic and Joule losses value was reduced in 1.37%.

Instituciones: Universidad Nacional de San Antonio Abad del Cusco