Long-Term Forecasting of Hydropower Generation and Electricity Demand using Artificial Neural Networks: Energy Gap Analysis between the Wlingi Hydropower Plant and Electricity Demand in Blitar Regency, Indonesia

Abstract

This study investigates the long-term structural electricity production–consumption gap under climate-induced hydrological variability, focusing on the Wlingi Hydroelectric Power Plant and Blitar Regency, Indonesia. Using historical operational and hydro-meteorological data from 2020–2025 as baseline input, three Artificial Neural Network (ANN) architectures Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRU) are comparatively evaluated to forecast electricity production, demand growth, and the resulting energy deficit over the 2026–2045 horizon. The predictive simulation reveals a progressively widening structural gap, expanding from approximately −427 GWh in 2026 to nearly −1,668 GWh in 2045, driven by exponential regional demand escalation and gradual hydropower performance degradation linked to global warming induced river discharge instability. While historically contributing 25%–37% of regional electricity demand, the hydropower fulfillment ratio is projected to decline to below 8% by 2045, indicating a substantial erosion of local renewable energy autonomy. Computational evaluation confirms that LSTM and GRU exhibit superior robustness and long-term dependency modeling compared to conventional RNN, particularly under stochastic hydro-climatic conditions. The novelty of this research lies in integrating climate-driven hydrological uncertainty into ANN based energy gap forecasting for operational hydropower systems. The findings underscore the vulnerability of climate-dependent renewable infrastructure and highlight the urgency of climate-adaptive energy diversification and storage integration strategies to sustain regional grid resilience under accelerating global warming pressures.