Time Series Analysis for Electricity Demand Forecasting: A Comparative Study of ARIMA and Exponential Smoothing Models in Indonesia

Abstract

The increasing global demand for electricity, driven by rapid urbanization and industrialization, necessitates accurate forecasting models to ensure efficient energy management. This study investigates electricity consumption patterns in Indonesia from 1970 to 2022 and evaluates time series forecasting methods for predicting future demand. The models employed include AutoRegressive Integrated Moving Average (ARIMA) and Exponential Smoothing, both of which are commonly used for short-term and long-term forecasts. The dataset was collected from Indonesia's national energy statistics, and preprocessing steps were applied to ensure data quality and consistency. Model performance was assessed using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). While ARIMA captured short-term trends, Exponential Smoothing demonstrated better long-term forecasting accuracy. The results highlight the effectiveness of these models in identifying electricity consumption trends and provide insights for policymakers and energy providers in optimizing energy distribution and production. Future work may incorporate advanced machine learning models and additional external factors for improved forecasting precision.