Weather and Air Quality Effects on Photovoltaic System Efficiency: A Modeling Approach

This paper aims to evaluate the performance of different machine learning algorithms for predicting the instantaneous efficiency of photovoltaic systems. The study utilized a dataset collected over seven months from seven distinct geographical locations characterized by diverse climatological conditions representative of the Colombian territory. The dataset included solar irradiance, climatological variables, air quality variables, and measured active power. Instantaneous efficiency was calculated from these data and defined as the study's target variable. An initial descriptive statistical analysis was performed to characterize the dataset. Subsequently, various prediction algorithms were executed and evaluated using standard error metrics. The evaluation demonstrated that incorporating air quality variables improved the predictive accuracy across all evaluated models. A key finding from the correlation analysis and feature importance assessment is the negative association between higher levels of environmental pollution and instantaneous efficiency.