Future Trends of Energy Consumption in Indonesia Through CNN-based Modeling

Accurate electricity load forecasting is essential for balancing supply and demand, optimizing energy distribution, and improving cost efficiency. This study presents a (CNN)based model for long-term electricity load forecasting in Indonesia, leveraging 1D Convolutional Layers (Conv1D) to extract patterns from historical consumption data (1989-2020) across multiple sectors. Using a sliding window approach, the model captures temporal dependencies and performs multi-step forecasting up to 30 years ahead. Evaluation against GRU and ARIMA shows CNN’s superior accuracy, achieving a MAPE of $4.30 \%$ and $\mathbf{R}^{\mathbf{2}}$ of $\mathbf{0. 8 7}$, outperforming GRU (MAPE 6.24%) and ARIMA (MAPE 8.60%). CNN’s ability to capture seasonal patterns and long-term trends without recurrent architectures makes it a promising tool for national energy planning. Its flexibility and scalability allow adaptation to various future scenarios, supporting sustainable electricity supply and informed policymaking.