Prediction and Comparative Analysis of Rooftop PV Solar Energy Efficiency Considering Indoor and Outdoor Parameters under Real Climate Conditions Factors with Machine Learning Model

Abstract

This research investigates the influence of indoor and outdoor factors on photovoltaic (PV) power generation at Utrecht University to accurately predict PV system performance by identifying critical impact factors and improving renewable energy efficiency. To predict plant efficiency, nineteen variables are analyzed, consisting of nine indoor photovoltaic panel characteristics (Open Circuit Voltage (Voc), Short Circuit Current (Isc), Maximum Power (Pmpp), Maximum Voltage (Umpp), Maximum Current (Impp), Filling Factor (FF), Parallel Resistance (Rp), Series Resistance (Rs), Module Temperature) and ten environmental factors (Air Temperature, Air Humidity, Dew Point, Air Pressure, Irradiation, Irradiation Propagation, Wind Speed, Wind Speed Propagation, Wind Direction, Wind Direction Propagation). This study provides a new perspective not previously addressed in the literature. In this study, different machine learning methods such as Multilayer Perceptron (MLP), Multivariate Adaptive Regression Spline (MARS), Multiple Linear Regression (MLR), and Random Forest (RF) models are used to predict power values using data from installed PV panels. Panel values obtained under real field conditions were used to train the models, and the results were compared. The Multilayer Perceptron (MLP) model was achieved with the highest classification accuracy of 0.990%. The machine learning models used for solar energy forecasting show high performance and produce results close to actual values. Models like Multi-Layer Perceptron (MLP) and Random Forest (RF) can be used in diverse locations based on load demand.