Predictive modelling with machine learning of garlic clove for novel designed solar drying system

Abstract

The design method of dual working medium, heat exchanger-evacuated tube assisted drying system is proposed in this work. Operation control method of solar drying system is specified. In order to achieve a drying cabin suitable air temperature controlling method and the expected operating environment of the system, the evacuated tube solar collector with dual-function hybrid drying system is considered as a case. Effect of operating ambient factors on efficiency of solar heating unit (SHU) is evaluated using traditional thermodynamics, as well as five different machine learning (ML) algorithms using Python software. The aim of this is to predict thermal performance of SHU. Influence of ambient temperature, solar intensity, and water flow in SHU on outlet air temperature as well as collecting efficiency are observed. Furthermore, ambient relative humidity is found to be the most important effect parameter for the outlet temperature of solar heating unit. As a result, the randomised lasso algorithm suggests that solar intensity is a crucial parameter. Gradient boosting regression tree (GBRT) can be considered as the good fit prediction performance with MAE of 0.31 and 0.59, RSME of 0.54 and 0.72, and coefficient of determination (R²) of 0.98 and 0.94 on training as well as testing values, correspondingly. This study also pointed out the direction for the application and development of thermodynamics analysis and machine learning.