Integrating computational models and machine learning for corrugated absorber plate solar collector thermal predictions

Abstract

This research provides a computational analysis, development, and empirical testing of a Corrugated Absorber Plate Solar Collector (CAPSC) with fins inserted Thermal Energy Storage (TES) system outlet temperature and thermal efficiency, which is then predicted employing machine learning techniques. For computational analysis, the governing equations of the CAPSC device have been generated and integrated into Computational Fluid Dynamics (CFD) software. Several experimental tests were conducted in both summer and winter seasons during the year, with varying mass flow rates ranging from 0.014 to 0.064 kg/s. The maximum average temperature difference among computational and experimental CAPSC outlet temperatures has been 1.72 °C at 0.044 kg/s. However, to enhance the thermal efficiency forecast of the CAPSC with fins inserted TES system, machine learning techniques such as Multiple Linear Regression (MLR), Random Forest (RF), K-Nearest Neighbors (KNN), eXtreme Gradient Boosting (XGBoost), and Artificial Neural Network (ANN) have been used. A substantial empirical dataset of 288 samples was utilized to analyze the machine-learning approaches. The observed statistical error and regression values in all techniques indicate the adequacy of the approaches presented. When compared to MLR, KNN, and RF techniques, ANN, followed by XGBoost, provides the most accurate forecast. Notably, XGBoost demonstrated superior accuracy due to its ability to handle complex non-linear relationships and interactions within the dataset. Furthermore, to improve the thermal efficiency forecast of the CAPSC with fins inserted TES system by integrating the computational model with the machine learning techniques by utilizing computational CAPSC output temperature as one of the inputs. This CFD with machine learning integration technique considerably enhanced the thermal performance forecast accuracy.