Assessing Machine Learning Models for Power Plant Generation Prediction

The difficulty of maintaining a balanced energy supply and demand has emerged as a key issue, creating a number of issues around the globe as a result of the rapidly increasing global population and industrial growth. This situation emphasizes the value of performing additional study on power generation and demand forecasts using machine learning approaches, particularly in Sri Lanka. In this analysis, we concentrate on projecting how much electricity two particular power plants, Laxapana and Mahaweli, will produce in the future. Our method involves parameterizing the net electricity generating data for each power plant as well as the hydro inflow data from multiple sub-power stations. We use a number of machine learning methods, such as Lasso Regression, Random Forest, and XGBoost. To further improve our models, we use methods like feature engineering and hyperparameter tuning with GridSearchCV and RandomizedSearchCV. In order to further enhance the predictive performance, we use model stacking.