Sistema elétrico para nanossatélites: rastreio da máxima potência através de redes neurais artificiais

• Ano de defesa: 2023
• Autor(a) principal: BATISTA, Leandro Souza
• Orientador(a): FONSECA NETO, João Viana da
• Banca de defesa: FONSECA NETO, Joao Viana da , SILVA, Luís Claudio de Oliveira , SOUZA, Francisco das Chagas de , MOURA, Jose Pinheiro de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Maranhão
• Programa de Pós-Graduação: PROGRAMA DE PÓS-GRADUAÇÃO EM ENGENHARIA AEROESPACIAL/CCET
• Departamento: DEPARTAMENTO DE ENGENHARIA DA ELETRICIDADE/CCET
• País: Brasil
• Palavras-chave em Português: EPS; nanossatélite; MPPT; redes neurais.
• Palavras-chave em Inglês: EPS; nanosatellite; MPPT; neural networks
• Área do conhecimento CNPq: Engenharia Aeroespacial; Engenharia Elétrica
• Link de acesso: https://tedebc.ufma.br/jspui/handle/tede/4788

The Electric Power System (EPS) is responsible for managing and distributing energy effi ciently to ensure the proper functioning of the nanosatellites. Therefore, this work presents an EPS project that addresses the monitoring of the maximum power point (MPPT) using Neural Networks. The efficiency of the method is demonstrated by comparing the Neural Network with other existing tracking objects such as Pertube & Observe and Incremental Conductance. Several configurations of neural networks were used, varying the number of neurons and their activation functions. Each network configuration goes through training, testing and validation steps using the MSE as the network stop configuration to select the best neural configuration. For training the neural network, the output data is used, the in cremental conductance signal that refers to the value of the PWM duty cycle that triggers the converter responsible for the MPPT. Solar irradiance and temperature were simulated by step functions, respectively. These values serve as input data for the neural network, as well as values corresponding to the load of the running subsystems. The tests were simu lated in a computational environment, where each of the three electrical systems contains an environment, Perbube & Observe, Incremental Conductance and neural networks. The MPP that thought the most neural network was faster and produced more electrical energy than other studies studied. Thus, choosing a neural network for MPPT becomes a good alternative to other maximum power tracking methods.