Developing machine learning models for relative humidity prediction in air-based energy systems and environmental management applications

• DOI: 10.1016/j.jenvman.2021.112736
• Publication Date: 2021-05-13T17:17:10Z
• Authors: Kinza Qadeer; Ashfaq Ahmad; Muhammad Abdul Qyyum; Abdul-Sattar Nizami; Moonyong Lee
• Keywords: Machine Learning; Conservation of Natural Resources; 13. Climate action; 0211 other engineering and technologies; Data Mining; Humidity; 02 engineering and technology; Algorithms

The prediction of relative humidity is a challenging task because of its nonlinear nature. The machine learning-based prediction strategies have attained significant attention in tackling a broad class of challenging nonlinear and complex problems. The random forest algorithm is a well-proven machine learning algorithm due to its ease of training and implementation, as it requires minimal preprocessing. The random forest algorithm has hitherto not been employed for estimating air quality parameters, such as relative humidity. In this study, the random forest approach is implemented to estimate the relative humidity as a function of dry- and wet-bulb temperatures. A well-known commercial process simulator called Aspen HYSYS® V10 is linked with MATLAB® version 2019a to establish a data mining environment. The robustness of the prediction model is evaluated against varying wet-bulb depressions. There is high absolute deviation that indicates a lower prediction performance of the model against the higher wet-bulb depression i.e., ~20.0 °C. The random forest model can predict relative humidity with a 1.1% mean absolute deviation compared to the values obtained through Aspen HYSYS. The performance of the RF estimation model is also compared with a well-known support vector regression model. The random forest model demonstrates 74.4% better performance than the support vector machine model for the problem of interest, i.e., relative humidity estimation. This study will significantly help the practitioners in efficient designing of air-dependent energy systems as well as in better environmental management through rigorous prediction of relative humidity.