A classifier to detect best mode for Solar Chimney Power Plant system

Abstract

Machine learning (ML) classifiers were used as a novel approach to select the best operating mode for Hybrid Solar Chimney Power Plant (HSCPP). The classifiers (decision tree (J48), Nave Bayes (NB), and Support Vector Machines (SVM)) were developed to identify the best operating modes of the HSCPP based on meteorological data sets. The HSCPP uses solar irradiation (SolarRad) to function as a power plant (PP) during the day and as a cooling tower (CT) at night. The SVM is the best classifier to predict the operating mode of HSCPP with an accuracy of ∼2% compared to NB and J48. Under the studied conditions the Regression analysis using REPTree was found to outperform SMOreg and achieved a relative absolute error ∼20 kWh. The productivity of the HSCPP is highly affected by maximum air temperature (Tair,Max), the average temperature of air (T air,Avg), solar irradiation standard deviation (SolarRadSTD), and maximum wind speed (Wsp,Max). Under optimal conditions, the HSCPP generates an additional 2.5% of energy equivalent to revenue of $3910.5 per year. Results show that ML can be used to optimize the operation of HSCPP for maximum electrical power and distilled water production.