Prediction of Hydropower Plant Electricity Production Dependence on Weather Conditions Using Machine Learning Approach

Abstract

To optimize the hydropower plant operations in the Sulawesi Generation Unit of PLN, this study proposes a data-driven approach to analyze electricity production by incorporating weather data. Utilizing historical data from January 2014 to December 2023, relevant indicators were extracted using machine learning algorithms. The integration of hydropower generation data, dam operational data, temperature, and rainfall enabled the prediction of electricity output through various models, including SARIMAX, Random Forest Regressor, Support Vector Regression, and Extreme Gradient Boosting. The dataset, consisting of 120 rows and 18 variables, demonstrated that combining diverse yet correlated data sources improve prediction accuracy. The best-performing model was validated and applied to forecast on new, unseen data. The findings indicate that machine learning offers a strategic advantage for PLN's decision-making in managing interconnected hydropower operations within the national power grid.