Metodologia de Otimização de Núcleo Ferromagnético de Reator por meio de Redes Neurais Artificiais

The optimization of electromagnetic devices such as transformers, reactors, filters and compensators is crucial for the efficiency and space utilization of an electrical system. In recent years, computational techniques such as Finite Element Method (FEM) and Machine Learning have been increasingly employed and achieving significant results. This paper presents a methodology that applies FEM and Artificial Neural Networks (ANN’s) to accelerate the brute force optimization of an iron core reactor. The methodology consists of modeling the equipment by FEM, simulating variations of parameters to form a database, training an ANN to learn to calculate the specific case under study within a range of input parameters, and then applying brute force optimization. The proposed methodology reduces the processing time of the optimization process, providing an efficient and accurate solution for the design of electromagnetic devices. The results show that the use of ANN’s improves the optimization process in terms of computational time and accuracy.