<![CDATA[Indonesia's rapid economic development and energy transition goals necessitate accurate long-term electricity demand forecasting to ensure supply security while optimizing infrastructure investments. This study addresses critical gaps in existing forecasting methodologies by developing a hybrid Grey Wolf Optimizer-Neural Network (GWO-NN) model specifically designed for emerging economy characteristics. While recent deep learning approaches (LSTM, CNN-LSTM) show promise for short-term forecasting, they often fail in long-term predictions due to limited adaptability to economic volatility and infrastructure constraints typical in developing nations. Our GWO-NN framework overcomes these limitations through intelligent hyperparameter optimization and multi-scenario modeling that captures Indonesia's unique socio-economic dynamics. The model incorporates 15 years of historical data (2010-2025) across seven key variables: GDP growth, population dynamics, temperature variations, industrial activity, urbanization rates, energy efficiency, and electrification progress. Rigorous validation against PLN's official projections reveals superior performance: Conservative scenario achieves exceptional 3.9% average absolute difference, Moderate scenario 19.0%, demonstrating significant improvement over traditional ARIMA models (35% error) and recent CNN-LSTM approaches (25% error). The 2034 demand projections range from 377.0 TWh (Conservative) to 546.1 TWh (Optimistic), providing policymakers with robust planning envelopes. This research contributes methodologically through hybrid metaheuristic optimization and practically through evidence-based planning support for Indonesia's renewable energy transition and carbon neutrality targets by 2060.]]>
