Electrical load forecasting in disaggregated levels using Fuzzy ARTMAP artificial neural network and noise removal by singular spectrum analysis

• Autor(a) principal: Müller, M. R. [UNESP]
• Data de Publicação: 2020
• Outros Autores: Gaio, G., Carreno, E. M., Lotufo, A. D.P. [UNESP], Teixeira, L. A.
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1007/s42452-020-2988-5; http://hdl.handle.net/11449/207031
• topic: Forecasting; Fuzzy ARTMAP; Power load; Singular spectrum analysis

Electrical load forecasting in disaggregated levels is a difficult task due to time series randomness, which leads to noise and consequently affects the quality of predictions. To mitigate this problem, noise removal using singular spectrum analysis (SSA) is used in this work in conjunction with a Fuzzy ARTMAP artificial neural network, presenting excellent results when compared with traditional methods like SARIMA. A reduction of almost 50% on the MAPE is achieved. The SSA method is preferable to other filtering methods because it has a low computational cost, depends on a small number of parameters, requires few data to present good results, and it does not cause delay into the denoised series.