Analysis of the importance of input data for short-term forecast of PV generation in an industrial plant

. The article presents a statistical analysis of data that constitute potential input variables for forecasting models in the task of predicting energy generation with a 15-minute horizon by a solar power plant operating for the needs of an industrial facility. A selection of these variables was made, followed by an evaluation of their importance for selected machine learning models. The forecasts used include: a Multilayer Perceptron neural network, random forest, gradient boosting decision trees and multiple linear regression. The article concludes with the findings from the conducted research.