Advancing Photovoltaic System Performance Analysis through Data Visualization and Intelligent Algorithms
Loading...
Date
2024
Journal Title
Journal ISSN
Volume Title
Publisher
National Higher School Of Technology And Engineering-ANNABA-
Abstract
Accurate prediction of the maximum power point (Pmpp) of photovoltaic (PV) systems is crucial to optimize the energy yield and maximizing the efficiency of solar energy harvesting.
This master thesis explores the potential of data-driven approaches for improving Pmpp prediction, utilizing regression techniques and feature importance analysis. Thestudyanalyzedadatasetofirradiance, temperature, andPmpp measurements, investigating the relationships between these variables and employing various regression models, including
Ridge Regression, Lasso Regression, Decision Tree Regression, and Random Forest Regression. Performance comparisons revealed that tree-based models, notably Random Forest, outperformed linear models in capturing the complex interactions between input features and Pmpp. Furthermore, feature importance analysis highlighted the significant influence of irradiance (GPOA)onPmpp, particularly for tree-based models, underscoring the need for accurate irradiance data and modeling techniques that effectively capture non-linear relationships. This master thesis concludes that data-driven approaches, specifically those employing tree-based models, hold significant potential for advancing Pmpp prediction and optimizing PV system performance. Future research should explore the integration of additional features, such as solar panel characteristics, atmospheric conditions, and system degradation factors, along with advanced machine learning techniques, to further enhance Pmpp prediction accuracy.