Previsão de demanda de cargas elétricas por seleção de variáveis stepwise e redes neurais artificiais

• Autor(a) principal: Alves, Marleide Ferreira [UNESP]
• Data de Publicação: 2013
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://hdl.handle.net/11449/111122
• topic: Carga e distribuição elétrica; Redes neurais (Computação); Algoritmos; Electric charge and distribution

With the increase in electric energy demand the planning of generation, transmission and distribution as well as the operation are important to provide services efficiently, economically and reliably. One of the tools to manage those resources are time series model forecasting. There are several models in the literature, as the regression models, statistical models, among others. Other model that has been highlighted in the literature is the forecasting using artificial neural network, due to the capacity of learning. Neural networks have several architectures, and one in particular, that is considered standard in the literature is the multilayer perceptron network with the backpropagation algorithm. The present work proposes a hybrid neural network composed by the linear regression method with stepwise variable selection with the multilayer perceptron artificial neural network with the backpropagation algorithm. The aim is to provide a simple and effective method to reduce the variables without losing the forecasting quality. The function of the linear regression model with stepwise variable selection is to select the more relevant variables to compose the input data set to training/ diagnostic of the multilayer perceptron neural network with the backpropagation algorithm that, consequently, is the responsible to realize the electric load forecasting. The aim of this proposal is to find a methodology that reduces the amount of input variables of the neural network and obtain satisfactory results. To verify the proposed methodology results are presented for electric short-term load forecasting in a period of 24 and 48 hours ahead, considering the historical data obtained from a company pertaining to the electrical sector