Prediction of energy production in a building-integrated photovoltaic system using machine learning algorithms

Abstract

Globally, the building sector is a significant consumer of energy. Implementing effective energy management system (EMS) strategies is crucial for reducing energy consumption and optimizing energy usage across all building elements. Moreover, solar energy is increasingly recognized as essential for sustainable building and urban development. Thus, the energy production forecast from photovoltaic systems installed on buildings is pivotal for a smart EMS's realization and optimal functioning. By incorporating accurate energy production forecasts, we can optimize energy consumption by fitting consumption patterns with production peaks and valleys. This study presents a comparative analysis of five widely used machine learning regressing models, including Random Forest, Decision Trees, XGBoost, LightGBM, and CatBoost for predicting the hourly energy production from a PV system mounted on a building. In this investigation, weather data and time-related characteristics were utilized as predictors. The results showed that LightGBM outperformed the other models, proving its suitability for PV energy forecasting tasks.