Drying characteristics, sustainability assessment and machine learning predictions of glass greenhouse drying system with photovoltaic roof

To enhance the applicability of photovoltaic drying systems in practical applications, a glass greenhouse drying system with photovoltaic roof system were proposed and tested by drying lemon experiment. The sustainability assessment and color analysis of the system was analyzed based on the environment-economic analysis. Furthermore, operation performances of the system were predicted by five machine learning methods. The results shown that the maximum electrical efficiency of this system was 23.80 %. Under the same outdoor conditions, the average drying rate of the proposed system and open sun drying were 0.2658 (g/g)/h and 0.3033 (g/g)/h, respectively. However, the lemon total color change value dried in the system with 3.15 has better than dried in the open sun drying with 4.28. The Prakash and Kumar model was founded to be the best kinetics model to describe lemon drying in the system. In addition, the economic payback period was 2.04 years, and the CO 2 mitigation was 6.33 ton/year of the drying system with a lifespan of 20 years. Finally, The Convolutional Neural Network (CNN) model was selected to be the optimal model to describe the temperature, humidity and power generation of the drying system with R 2 value 0.9917, 0.9928 and 0.9937, respectively.