Forecasting of electrical energy consumption using Autoregressive Integrated Moving Average (Case Study: ULP Meulaboh Kota)

Abstract

Forecasting electricity consumption is one of the solutions that can be implemented by the ULP Meulaboh Kota to ensure the availability of sufficient electricity supply. With the continuous increase in electricity demand, the ULP faces challenges in predicting and managing electricity consumption. Uncertainty in consumption patterns can lead to imbalances between supply and demand, potentially causing various issues such as power outages, high operational costs, and customer dissatisfaction. Therefore, accurate forecasting is essential to support effective decision-making and planning. This study aims to forecast electricity consumption across five different sectors: residential, social, business, industrial, and public, using the ARIMA (Autoregressive Integrated Moving Average) method. The forecasting process involves data collection, stationarity testing using the Augmented Dickey-Fuller (ADF) test, and differencing when necessary to achieve stationarity. The ARIMA model is identified through ACF and PACF plot analysis, estimated, and tested before being used for forecasting. The results indicate that the ARIMA method provides highly accurate forecasts for all sectors, as reflected by the low Mean Absolute Percentage Error (MAPE) values. The residential sector has a MAPE of 4.3957%, the social sector 4.3757%, the business sector 3.1125%, the industrial sector 7.9937%, and the public sector 4.3646%. Overall, the forecasting error produced by the ARIMA model remains below 8%, with an average MAPE of 4.8483% across all sectors.