Hydroelectricity Generation Forecast using an Autoregressive Integrated Moving Average Model

Abstract

Hydroelectricity is the most efficient method for producing large amounts of electricity. Therefore, accurate hydroelectricity generation forecast is critical to the power system's reliable, secure, safe, and cost-effective operation. Meanwhile, most of the studies rely solely on multivariate techniques that employ hydrological and metrological variables characterized by complex nonlinearities, thereby making the forecast more difficult and prone to errors. Hence, there is a need to explore univariate techniques such as an autoregressive integrated moving average (ARIMA) for a reliable and accurate forecast. In this study, Kainji hydroelectricity generation was forecasted using an ARIMA model. The forecast period, which ran from January 2023 to December 2033, was based on data collected from January 2010 to December 2022. The ARIMA models were implemented using a practical Box-Jenkins method that consists of four iterative sequences: stationarity check, model identification, estimation of parameters, and diagnostics. To make the forecast, an ARIMA (12, 1, 16) was used, since it performed better than other prospective models in terms of the adjusted determination coefficient, Akaike Information Criterion, Swartz Bayesian Information Criterion, and Hannan-Quinn Information Criterion. Cumulative electricity of 48,466.4 MWh is projected to be produced over the next 10 years. It is hoped that the study's findings will facilitate the planning and allocation of resources for seamless operation of the station.