Comparative Analysis of Machine Learning Models for Day-Ahead Photovoltaic Power Production Forecasting

A main challenge for integrating the intermittent photovoltaic (PV) power generation remains accuracy of day-ahead forecasts and establishment robust performing methods. The purpose this work is to address these technological challenges by evaluating PV production forecasting performance different machine learning models under supervised regimes minimal input features. Specifically, capability Bayesian neural network (BNN), support vector regression (SVR), tree (RT) was investigated employing same dataset training verification, thus enabling a valid comparison. regime analysis demonstrated that strongly dependent on timeframe train set, data sequence, application irradiance condition filters. Furthermore, accurate results were obtained utilizing only measured output other calculated parameters training. Consequently, useful information provided establishing methodology utilizes an optimal approach. Finally, optimally constructed BNN outperformed all achieving accuracies lower than 5%.