A hybrid CFD and machine learning study of energy performance of photovoltaic systems with a porous collector: Model development and validation

This study investigates the predictive modeling of temperature (T(K)) using a dataset of over 128,000 data points characterized by x, y, and z coordinates as inputs. The case study considered here is a photovoltaic system with porous collector for enhancing the efficiency of solar system. Computatio...

Full description

Saved in:
Bibliographic Details
Main Authors: Yinling Wang, Lei Yu, Mazhar Ali, Imran Ali Khan, Tahir Maqsood, Haining Gao, Qi Wang, Xiaolei Guo
Format: Article
Language:English
Published: Elsevier 2025-05-01
Series:Case Studies in Thermal Engineering
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2214157X25002588
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study investigates the predictive modeling of temperature (T(K)) using a dataset of over 128,000 data points characterized by x, y, and z coordinates as inputs. The case study considered here is a photovoltaic system with porous collector for enhancing the efficiency of solar system. Computational modeling was carried out via CFD (Computational Fluid Dynamics), and the temperature distribution was determined which was later used in machine learning (ML) evaluation. Indeed, a hybrid model was developed combining CFD and ML for the first time to predict temperature distribution versus special coordinates in a photovoltaic thermal system. Three advanced machine learning models, i.e., Gradient Boosting (GB), Extreme Gradient Boosting (XGB), and Histogram-based Gradient Boosting (HGB) were applied to analyze and predict T in system. A systematic preprocessing pipeline was developed to enhance model performance, including outlier detection and feature normalization. Hyperparameter optimization process in this study uses the Water Cycle Algorithm (WCA), a metaheuristic method inspired by natural processes. Among the models, XGB emerged as the best performer, revealing a total R2 of 0.99823, a Root Mean Square Error (RMSE) of 0.06596, and a Mean Absolute Error (MAE) of 0.04442. These results demonstrated the capability of machine learning to accurately capture complex relationships within structured datasets.
ISSN:2214-157X