Applicability of levenberg-marquardt algorithm for power generation analysis of the system photovoltaic

• Autor(a) principal: Pinheiro, Elisangela
• Data de Publicação: 2017
• Outros Autores: Ruther, Ricardo, Lovato, Adalberto
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Revista Produção Online
• Texto Completo: https://www.producaoonline.org.br/rpo/article/view/2542
• topic: Photovoltaic Solar Energy. Artificial Neural Networks. Levemberg– Marquardt Algorithm. Transfer function logsigmoide. Transfer function purelin.; Energia solar fotovoltaica. Redes Neurais Artificiais. Algoritmo Levemberg-Marquardt. Função de transferência Logsigmóide. Função de transferência purelin.

This paper talk about the applicability of artificial neural networks for power generation analysis for a photovoltaic system connected to the grid. First, the characteristics and mathematical potential of data processing presented by networks are described. Then, several configurations are tested in order to find the most appropriate one. Data from 2014 were used for energy generation, room temperature, module temperature, solar radiation incidence and time of day. The application used the Levenberg-Marquardt algorithms to obtain the network parameters and the criterion of the mean square error to measure the performance. The training was reealized with 5, 10, 15, 20, 25, 30 and 60 neurons in the hidden layer. The transfer functions were logsigmoid and purelin. The best result was obtained with 25 neurons, with a correlation coefficient of 0.98.