Transformer-based artificial intelligence for forecasting energy demand in irrigation districts

• Develop high-accuracy forecasting model for mid-term hourly irrigation energy demand. • Identify 12 key variables impacting irrigation energy consumption patterns. • Apply multi-head attention mechanism to enhance model interpretability and accuracy. • Optimize energy use, support sustainable operations, and reduce operational costs. Efficient energy management in pressurized irrigation systems is essential to optimize water and energy use. This study presents a novel hybrid forecasting model based on Transformer Neural Networks (TNNs) integrated with fuzzy logic (FL) for mid-term hourly energy demand prediction in irrigation districts (ID). The model was applied to the Valle Inferior del Guadalquivir Irrigation District (VIGID), Spain, using real data from 2020 to 2023. A total of 26 potential input variables were initially considered, including climatic conditions, energy prices, crop distribution, and historical energy demand ( ED ). Through a rigorous statistical analysis combining FL and correlation matrices, 12 key variables were selected to capture the most relevant temporal and meteorological influences. A sequence-to-sequence architecture was implemented to model complex temporal dependencies in ED patterns. The model achieved high predictive performance, with an average coefficient of determination (R 2 ) of 99.62%, Root Mean Squared Error (RMSE) of 0.038 and Mean Absolute Error (MAE) of 32.36 kWh, demonstrating its ability to capture nonlinear behaviour and seasonal variability. The multi-head attention mechanism enhanced interpretability by dynamically weighting inputs according to context, while FL contributed to robustness under uncertainty. Results confirm the effectiveness of attention-based architectures for irrigation energy forecasting, supporting data-driven decision-making. Accurate ED prediction is key to optimizing resource allocation, reducing operational costs, and facilitating the integration of renewable energy sources into agricultural systems. This research underscores the potential of AI-driven tools to enhance energy efficiency and promote sustainability in modern irrigation practices.