Predictive Modeling of Shading Effects on Photovoltaic Panels Using Regression Analysis

TL;DR

This paper evaluates five regression models to predict shading effects on photovoltaic panels, demonstrating that XGBoost and random forest models achieve high accuracy, which can enhance PV-powered drone efficiency and disaster response capabilities.

Contribution

It introduces a comparative analysis of multiple regression models for shading prediction on PV panels, highlighting the superior performance of XGBoost and random forest methods.

Findings

XGBoost achieved an R2 score of 0.926.

Random forest regression also performed well.

Accurate shading prediction can improve drone PV panel efficiency.

Abstract

Drones have become indispensable assets during human-made and natural disasters, offering damage assessment, aid delivery, and communication restoration capabilities. However, most drones rely on batteries that require frequent recharging, limiting their effectiveness in continuous missions. Photovoltaic (PV) powered drones are an ideal alternative. However, their performance degrades in variable lighting conditions. Hence, machine learning (ML) controlled PV cells present a promising solution for extending the endurance of a drone. This work evaluates five regression models, linear regression, lasso regression, ridge regression, random forest regression, and XGBoost regression, to predict shading percentages on PV panels. Accurate prediction of shading is crucial for improving the performance and efficiency of ML-controlled PV panels in varying conditions. By achieving a lower MSE and higher R2 Scores, XGBoost and random forest methods were the best-performing regression models. Notably, XGBoost showed superior performance with an R2 Score of 0.926. These findings highlight the possibility of utilizing the regression model to enhance PV-powered drones' efficiency, prolong flight time, reduce maintenance costs, and improve disaster response capabilities.