Improving the thermal efficiency of a solar flat plate collector using MWCNT-Fe3O4/water hybrid nanofluids and ensemble machine learning

The thermal performance of a flat plate solar collector using MWCNT + Fe3O4/Water hybrid nanofluids was examined in this research. The flat plate solar collector was tested using different nanofluid concentrations and flow rates in an arid environment. A significant enhancement in coefficient of heat transfer (26.3%) with a marginal loss on pressure drop due to friction factor (18.9%). The data collected during experimental testing was utilized to develop novel prediction models for efficient heat transfer, Nusselt's number, friction factor, and thermal efficiency. The modern ensemble machine learning techniques Boosted Regression Tree (BRT) and Extreme Gradient Boosting (XGBoost) were used to develop prognostic models for each parameter. A battery of statistical methods and Taylor's graphs were used to compare the performance of these two modern ML techniques. The value of R2 for the BRT-based prediction models were 0.9619 -0.9994 and 0.9914 -0.9997 for XGBoost-based models. The mean squared error was quite low for all the models (0.000081 -9.11), while the mean absolute percentage error was negligible from 0.0025 to 0.3114. The comprehensive statistical analysis of the prognostic model was com-plemented with Taylor's graphs to develop an improved comparison paradigm, to reveal the superiority of XGBoost over BRT.

Automated summary

The thermal performance of a flat plate solar collector using MWCNT + Fe3O4/Water hybrid nanofluids was examined, and the impact of nanofluid concentrations and flow rates on heat transfer coefficient and pressure drop was investigated. Prediction models were developed to improve heat transfer, Nusselt's number, friction factor, and thermal efficiency using Boosted Regression Tree (BRT) and Extreme Gradient Boosting (XGBoost) techniques. The statistical analysis and comparison showed the superiority of XGBoost over BRT.