Hybrid Multi-Stage Forecasting for Sustainable Electricity Demand Planning in the Netherlands

Accurate forecasting is essential for effective energy management, particularly in evolving and data-driven energy markets. This study presents a hybrid multi-stage forecasting framework to enhance short- and long-term electricity demand predictions in the Netherlands. We compare statistical (SARIMAX), hybrid (SARIMAX-LSTM), and deep learning (sequence-to-sequence) models across forecasting horizons ranging from 1 to 180 days. The models were trained on daily load data from ENTSO-E (2009-2023) and enriched with external variables, including weather conditions, energy prices, socioeconomic indicators, and engineered features such as calendar effects and rolling demand windows. Three feature configurations were evaluated: exogenous-only, generated-only, and a combined set. Internally generated features consistently outperformed exogenous inputs, particularly in long-term forecasts. The sequence-to-sequence model achieved the highest accuracy at the 180-day horizon, with a mean absolute percentage error (MAPE) of approximately 1.88%, outperforming both SARIMAX and the SARIMAX-LSTM hybrid. An additional SARIMAX-based analysis assesses the individual effects of renewable and socioeconomic indicators. Renewable energy production data improve short-term accuracy (MAPE reduced from 2.13% to 1.09%), but contribute little to long-term predictions. Socioeconomic variables slightly degrade predictive performance.