Comparing Machine Learning and Neuro-Fuzzy Models for Hydropower Prediction Under Seasonal Operating Conditions

Abstract

Accurate prediction of hydropower energy production is essential for operational planning and decision-making under varying hydrological conditions. However, comparative evaluations of high-accuracy data-driven models and interpretable intelligent models using real operational data remain limited. This study aims to compare the performance of Machine Learning and Neuro-Fuzzy approaches for predicting energy production at the Sengguruh Hydropower Plant, Indonesia. One year of engineering-validated operational data consisting of discharge, reservoir and tailrace water levels, operation time, turbine efficiency, and derived hydraulic head were used for model development and evaluation. Three Machine Learning models, namely Artificial Neural Network (ANN), Random Forest (RF), and Support Vector Regression (SVR), were compared with an Adaptive Neuro-Fuzzy Inference System (ANFIS). Model performance was evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). The results indicate that Machine Learning models achieved superior predictive performance, with Random Forest providing the highest accuracy (RMSE = 8.42 MWh/day, MAE = 6.35 MWh/day, and R² = 0.982). In contrast, the ANFIS model produced lower prediction accuracy but offered a more interpretable representation of hydropower operating behavior. Correlation analysis further indicates that operation time and inflow discharge are the most influential operational variables associated with energy production. The findings demonstrate that Machine Learning approaches are more suitable for high-accuracy forecasting and operational optimization, whereas Neuro-Fuzzy models provide advantages in interpretability and operational representation. These results contribute to the development of intelligent hydropower forecasting systems that balance predictive performance and engineering understanding.