Data driven insights for parabolic trough solar collectors: Artificial intelligence-based energy and exergy performance analysis

Abstract

Artificial intelligence (AI) algorithms can potentially contribute to optimizing energy and exergy outputs in renewable resources to increase efficiencies and reduce environmental risk. This study utilized tree-based, linear, and non-linear regression techniques to predict the energy and exergy efficiency of Parabolic Trough Solar Collectors (PTSCs) using oil-based nanofluids. The cooling fluids were prepared from three main oil types, namely Therminol VP-1, Syltherm 800, and Dowtherm Q mixed with three metallic oxides, including Al2O3, CuO, and SiO2, in various volume fractions. The two outputs were predicted according to a range of input parameters, namely Volume Fraction (%), Reynolds Number (Re), Inlet Fluid Temperature, Direct Solar Irradiance, Nusselt Number (Nu), and Friction Factor (f). Ensemble approaches such as Extra Trees Regressor (ETR), Extreme Gradient Boosting (XGBoost), Random Forest Regressor (RFR), Classification and Regression Trees (CART), and Adaptive Boosting (AdaBoost) were the top-performing models in the model selection process out of nine. The modeling results showed that, CART was the top model in predicting the energy efficiency using Syltherm 800-SiO2 nanofluid with R2 = 0.9999. Meanwhile, ETR was the top model in predicting the exergy efficiency using Dowtherm Q-SiO2 nanofluid with R2 = 0.9988. Moreover, in the business insights, the maximum errors in the energy and exergy models were observed (1.43 % and 1.97 %) using Therminol VP-1, (1.3 % and 2.44 %) using Syltherm 800 and Syltherm 800-CuO and (1.15 % and 2 %) using Dowtherm Q and Dowtherm Q-CuO, respectively.