PCA-Kalman based load forecasting of electric power demand

Abstract

Electricity demand time series are stochastic processes related to climate, social and economic variables. By predicting the evolution of such time series, electrical load forecasting can be performed in order to support the electrical grid planning. In this paper, we propose a Kalman based load forecasting system for daily demand forecasting. Our proposed approach incorporates a Principal Component Analysis (PCA) of the input variables obtained from linear and nonlinear transformations of the candidate time series. In order to validate our predicting scheme, data collected from Brasília distribution company has been used. Our proposed approach outperforms state-of-the-art approaches based on state space and artificial neural networks.