Predictive assessment from ANN and MLR models to optimize the ideal evaporative/hybrid cooler based on experimental observations

Abstract

Single-stage evaporative coolers like a direct evaporative cooler (DEC) and indirect evaporative cooler (IEC) are not versatile in their applicability in diverse climatic conditions due to their dependency on local climatic variation. To overcome this, an indigenously designed three-stage cooling system capable of operating in single-stage modes like DEC, IEC, and direct expansion (DX), and multi-stage modes like IEC-DEC, DEC-DX, and IEC-DEC-DX is developed. Detailed energetic and exergetic investigations are conducted based on experimental data acquired from 20th-26th of each month from April to July 2019. The thermodynamic analyses lead to optimizing the ideal cooling system for tropical climate through the development and comparison of artificial neural network (ANN) and multiple linear regression (MLR) models by predicting the wet-bulb effectiveness (ε), coefficient of performance (COP), and exergy efficiency (ηex). The input parameters for the predictive models include ambient temperature, ambient relative humidity, conditioned air temperature, and corresponding relative humidity. Based on the predictions, it can be concluded that the ANN is superior to the MLR method for performance prediction, with a minimum mean square error. Although the IEC-DEC is observed ideal for tropical climate due to higher cooling capacity and COP in extreme summer, the efficacy of IEC-DEC-DX cannot be ignored for its all-weather competency. The ANN models predict thermodynamic performance parameters of proposed coolers within the error limit of ±10%.