Electricity demand prediction for sustainable development in Cambodia using recurrent neural networks with ERA5 reanalysis climate variables

Sustainable energy development plays a prominent role in energy planning to maintain natural resources and mitigate the usage of fossil fuels. The atmospheric factor is one of the main influencing factors that changed the electricity consumption's behavior due to global warming. In this study, the recurrent neural network (RNN) models were developed to examine the effects of 66 climate variables, collected from the European Center for Medium-Range Weather Forecast (ECMWF) ERA5 reanalysis, on power demand in Cambodia. The statistically significant climate variables were filtered by considering the cross-correlation between power demand and each climate variable. Moreover, the wide range of feedback delays was computed from the power demand dataset and was defined using the 95% confidence intervals. The comparison between a nonlinear autoregressive neural network with exogenous inputs (NARX) using historical power demand with the correlated climate variables and a nonlinear autoregressive neural network (NAR) using only historical power demand dataset was made. The various benchmarked models were evaluated and compared for their performances using statistical indices such as normalized root-mean-square error (NMSE) and coefficient of determination (R-2). The results showed the NARX model could perform better than the NAR model for predicting electricity demand time-series. (C) 2022 The Author(s). Published by Elsevier Ltd.