Statistical modeling of photovoltaic performance via PCA and regression: a case study

Abstract

The performance of photovoltaic (PV) systems is highly influenced by a complex set of climatic and industrial parameters. This study aims to identify the most significant factors affecting PV power output by combining a predictive modeling approach using multiple linear regression with a multivariate statistical analysis based on Principal Component Analysis (PCA). The analysis is based on real-world weekly data that includes a wide range of \textbf{climatic factors} (solar irradiance, ambient and cell temperature, relative humidity, wind speed) and \textbf{industrial or technological parameters} (bandgap energy, diode ideality factor, thermal coefficient, angle of incidence ). The linear regression model quantifies the individual impact of each variable on PV power, while PCA reveals the latent structure of the dataset by identifying the most informative linear combinations of variables. Results consistently highlight solar irradiance, cell temperature, and technological factors such as $E_g$ and $A$ as the most critical determinants of PV output. This combined approach provides a robust and interpretable method for reducing model complexity, with practical implications for the design, monitoring, and optimization of photovoltaic systems in real-world conditions.