4 - Forecasting of photovoltaic output power using machine learning and deep learning algorithms

Abstract

In this chapter, numerous examples of the application of machine learning and deep learning algorithms for short-term photovoltaic output power forecasting are presented and discussed. The first application is about forecasting one-step ahead photovoltaic output power, using machine learning algorithms; in this application, only historical data are used. The second application focuses on the use of artificial neural networks to forecast photovoltaic output power using measured photovoltaic power and other meteorological parameters. The third application concerns the investigation of ensemble learning methods for one-step ahead forecasting. The fourth application deals with the utilization of deep learning natural networks for one-step and multistep ahead forecasting of output photovoltaic power. The investigated algorithms include deep neural networks like LSTM, bidirectional LSTM, GRU, CNN, and hybrid configurations. The fifth application shows how to qualify variation in the predicted results by calculating the confidence interval and the prediction interval, which is very important for analyzing the uncertainties of the forecasters.