Forecasting electricity demand in Japan: A Bayesian spatial autoregressive ARMA approach

It was confirmed that the space-time model improves the performance of forecasting future electricity demand in Japan and the spatial autoregressive ARMA (1, 1) model was selected.

Abstract

Regional electricity demand in Japan and spatial interaction among the regions using a Bayesian approach were examined. A spatial autoregressive (SAR) ARMA model was proposed to consider the features of electricity demand in Japan and a strategy of Markov chain Monte Carlo (MCMC) methods was constructed to estimate the parameters of the model. From empirical results, the spatial autoregressive ARMA (1, 1) model was selected, and it was found that spatial interaction plays an important role in electricity demand in Japan. Moreover, log predictive density showed that this SAR-ARMA model performs better than a univariate ARMA model. It was confirmed that the space–time model improves the performance of forecasting future electricity demand in Japan.