Performance Modeling of Discharging Process in a Thermochemical Fluid System Using Machine Learning Approaches

Open absorption system is an energy efficient method for dehumidification/humidification. As the global demand for clean energy continues to rise, the significance of machine learning in forecasting energy systems will amplify. In this paper, we developed three machine learning algorithms to accurately forecast air output mixing ratio and temperature of a thermochemical fluid system and compare their performance. The variables used to forecast our responses are ambient air temperature (°C), inlet air mixing ratio (g/kg), the inlet desiccant temperature(°C), the air mass flow rate (kg/s) and desiccant mass flow rate (kg/s). The evaluation metrics shows Random Forest regressor greatly outperform multi-layer perceptron neural network, and is slightly better than Gradient boosting regressor. The Random Forest regressor stands out with a significant R2 of 0.999 for predicting both air output mixing ratio and temperature, accompanied by MAPE values of 0.241 and 0.097, for air output mixing ratio and temperature prediction, respectively.