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

Area of Science:

• Environmental Science
• Data Science
• Machine Learning

Background:

• Relative humidity prediction is complex due to its nonlinear behavior.
• Machine learning offers powerful solutions for nonlinear and complex problems.
• The random forest algorithm is effective, requiring minimal preprocessing.

Purpose of the Study:

• To implement the random forest approach for relative humidity estimation.
• To evaluate the model's performance against varying wet-bulb depressions.
• To compare the random forest model with a support vector regression model.

Main Methods:

• Utilized the random forest algorithm to estimate relative humidity based on dry- and wet-bulb temperatures.
• Integrated Aspen HYSYS V10 with MATLAB 2019a for a data mining environment.
• Assessed model robustness using varying wet-bulb depressions and compared with support vector regression.

Main Results:

• The random forest model achieved a 1.1% mean absolute deviation in relative humidity prediction compared to Aspen HYSYS.
• Model performance decreased with higher wet-bulb depressions (around 20.0 °C).
• The random forest model outperformed the support vector regression model by 74.4%.

Conclusions:

• The random forest approach provides a robust and accurate method for relative humidity estimation.
• This technique offers significant benefits for designing air-dependent energy systems and environmental management.
• Further research may explore model performance at higher wet-bulb depressions.