Título: Monthly Electric Energy Demand Forecasting By Fuzzy Inference System
Autores: Luna, Ivette; Ballini, Rosangela
Resumo: This paper presents the use of a fuzzy rule-based system for mid-term electric energy demand forecasting. The results are achieved for an specific region at the southeastern part of Brazil. The total demand is divided in three groups of consuption: residential, industrial and commercial. The forecasting model adopted is based on Takagi-Sugeno fuzzy rules, where the number of fuzzy rules is defined by the Subtractive Clustering algorithm, an unsupervised approach applied over an in-sample data set. A fuzzy rule base is determined by each group of consumption and the model parameters are adjusted using the Expectation-Maximization optimization algorithm. As input variables are considered the observations of demand in previous moments as well as macroeconomic explanatory variables. Forecasting tests over in-sample and out-of-sample data sets are developed. The results show the adequacy of the models adjusted, achieving annual absolute percentage errors of 3% in average.
Palavras-chave: Electric energy demand; fuzzy inference system; forecasting; time series
Páginas: 8
Código DOI: 10.21528/lmln-vol10-no2-art6
Artigo em PDF: vol10-no2-art6.pdf
Arquivo BibTex: vol10-no2-art6.bib