Machine learning techniques validation using experimental average surface temperature and output power of photovoltaic panel cooled by porous media under indoor condition

The main objective of this paper is to increase electrical efficiency by cooling photovoltaic cells with porous materials and using different machine learning models to predict the photovoltaic panel average surface temperature and output power that is cooled by porous media immersed in water, and then validate the machine learning models result with the experimental results. Six different models of machine learning were studied, namely; AdaBoost, Random Forest, Tree, Support Vector Machine, Gradient Boosting, and Linear Regression to find out the impact factor of various photovoltaic panel cooling parameters on the photovoltaic panel power output and the efficiency. The input parameters were ambient temperature, solar radiation, inlet water temperature, porosity size, water volume flow, and time. The experimental results showed that the best-used porosity size was 0.35. Predictions from the machine learning models on the photovoltaic panel cooling found that the photovoltaic panel average surface temperature, and the performance of algorithm AdaBoost were closer to the experimental results followed by the values of root mean square error of 0.5135, mean absolute error of 0.3829 and R2 of 0.986. While the models of the algorithm Linear Regression, has the values of Root Means Square Error of 1.1756, mean absolute error of 0.8673, and R2 of 0.929 respectively are far away from the experimental results. Therefore, the algorithm AdaBoost model prediction results proved one of the best machine learning models for analytical studies on photovoltaic panel cooling performance predictions. Also, the second best-performed model is the algorithm SVM, having low error values of RMSE 0.1057, MAE 0.0889, and R2 0.9892 respectively, which can be used for validation purposes as well. Moreover, the analyses by the machine learning model of the algorithm AdaBoost using machine learning techniques showed that the porosity size plays the most effective parameter on the photovoltaic panel average temperature and power output are similar to the findings of the experimental results.

Text 124777.pdf - Published Version Restricted to Repository staff only Download (1MB) Official URL or Download Paper: https://www.praiseworthyprize.org/jsm/index.php?jo...

Actions (login required)

View Item