Computational Model for Photovoltaic Solar Energy Forecasting Based on the K-Nearest Neighbor Method

Abstract

Integrating PV technologies into power systems requires precise planning of PV performance. The ability to predict solar photovoltaic generation is a challenge for its integration into electrical systems. Improvements in forecasting models with more accurate results and fewer errors are necessary for the future development of microgrid projects and the dispatch of the economic sector. This research presents a computational machine learning model to predict the PV output power using historical PV output power data from a 960 kWP grid-connected PV system in southern Italy. The results showed agreement between the predicted and actual values, with errors ranging from 5% to 12%. We concluded that using machine learning techniques makes it feasible to predict the photovoltaic output power.