Forecasting Energy Consumption on a Microgrid using ARIMA-GRU Model

Abstract

Accurate forecasting of energy consumption is crucial for the efficient management and control of modern energy grids, particularly amid the escalating integration of renewable energy sources. This study proposes a hybrid approach that combines the Autoregressive Integrated Moving Average (ARIMA) and the Gated Recurrent Unit neural network (GRU) to predict energy consumption in a microgrid setting. The proposed hybrid ARIMA-GRU model integrates ARIMA’s residuals with GRU’s non-linear modeling capabilities, enabling enhanced prediction accuracy while capturing both linear and non-linear dependencies in microgrid energy data. The model’s performance is evaluated using real-world energy consumption data, achieving an RMSE of 38.28 kWh, MAE of 31.24 kWh, and MAPE of 10.29%. These results highlight the model’s effectiveness in improving energy forecasting and providing practical insights for better energy management in microgrids.