Abstract
Electricity consumption exhibits a sustained upward trend driven by population growth, industrial expansion and increasing household demand for electrical appliances. As urban development intensifies, accurate electricity demand prediction becomes essential to support capacity planning and ensure a reliable and sustainable power distribution system. This study aims to enhance the accuracy of monthly electricity consumption prediction through a comparative analysis of three time series models, namely the Autoregressive Integrated Moving Average (ARIMA), Autoregressive Integrated Moving Average with Exogenous Variables (ARIMAX) and Autoregressive Fractionally Integrated Moving Average (ARFIMA) models. Historical monthly electricity consumption data, incorporating the number of working days as an exogenous variable, were analyzed through data transformation, stationarity testing and model identification stages. Model performance was evaluated using statistical indicators, including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). The results indicate that the ARIMA (4,1,4), ARIMAX (2,1,4) and ARFIMA (4,0.2996,4) models satisfy the AIC and BIC criteria; however, based on RMSE, MAE and MAPE, the ARIMAX (2,1,4) model provides the highest prediction accuracy. The inclusion of the number of working days as an exogenous variable significantly improves predictive performance compared to ARIMA and ARFIMA. These findings underscore the importance of incorporating relevant external factors to enhance electricity demand prediction and offer practical implications for electricity utilities and local governments in demand planning, distribution optimization and sustainable energy policy formulation.
Keywords: ARIMAX, Electricity Consumption, Exogenous Variables, Model Comparison, Time Series Models.