Solar Power Forecasting using Stacking Ensemble Models with Bayesian Meta-Learning

Predicting solar power generation is vital for better uses of renewable energy farms. This paper proposes averaging and stacking ensemble models for predicting solar power generation. The machine learning (ML) models include Least Absolute Shrinkage and Selection Operator (LASSO) regression, random forest (RF), multilayer perceptron (MLP), support vector machine regression (SVR), and extreme gradient boosting (XGB). In order to improve the performance of the models, two ensemble methods, namely averaging ensemble with linear weighted average and stacking ensemble models, are developed. The models are validated on real-time series of historical datasets from the Asian Institute of Technology (AIT), Thailand. The analysis shows that the stacking ensemble model with Bayesian as meta-learner achieves better prediction accuracy than all tested models.