Time series modeling and forecasting of total primary energy consumption in Sri Lanka

Abstract

Primary energy is the energy that is harvested directly from natural resources. Forecasting total primary energy consumption in Sri Lanka is significant as primary energy consumption worldwide is expected to continue increasing. This study aimed to model and forecast total primary energy consumption in Sri Lanka, which has not yet been analysed using Time Series Analysis. For this purpose, the annual data of total primary energy consumption in Sri Lanka from 1960 to 2019 in terawatt-hours was extracted from the world wide web and analysed with Auto- Regressive Integrated Moving-Average (ARIMA) model. The stationary of the series was tested using the Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test, Phillips-Perron (PP) test, and Augmented Dickey-Fuller (ADF) test. The study revealed the ARIMA(4,2,1) model as a best- fitting model, which gave the minimum value of Akaike Information Criterion (AIC). Total primary energy consumption from 2008 to 2019 was forecasted using ARIMA(4,2,1) model as it satisfied the model diagnostics, which are ARCH test, autocorrelation function, and normality of residuals. With Mean Absolute Error (MAE) of 5.0283 and Root Mean Squared Error (RMSE) of 5.9216, the results illustrate that ARIMA(4,2,1) model captures the trend in total primary energy consumption accurately. Based on the results, the study suggests ARIMA(4,2,1) is more convenient in determining the trends and the patterns of the future in total primary energy consumption in Sri Lanka.