Machine Learning-Based Prediction and Optimization Model for Indirect Evaporative Cooling Units

Rapid performance evaluation of evaporative cooling systems is of great significance for assessing their suitability and facilitating widespread implementation. This study employs various machine learning models to predict the wet-bulb efficiency of an indirect evaporative cooling unit, with the trained models subsequently applied to predictions in other regions. Experimental tests were conducted to collect operational parameters such as the dry-bulb temperature, wet-bulb temperature, flow rate, and secondary/primary air volume ratio of the secondary air at different time periods. Principal component analysis (PCA) was used to identify the key operational parameters that significantly influence wet-bulb efficiency. The results show that the Relevance Vector Machine (RVM) model achieves good agreement with experimental values: for the test set, the correlation coefficient R2 is 0.9456, RMSE is 0.02452, and MAPE is 3.6748%; for the validation set, R2 is 0.8338, RMSE is 0.04318, and MAPE is 6.5069%. The trained model was also used to predict the unit’s wet-bulb efficiency in Beijing, Guangzhou, Xi’an, and Lanzhou. This study provides a comparison of multiple feasible methods for predicting and optimizing the wet-bulb efficiency of indirect evaporative cooling units, offering valuable insights for the design optimization and rapid performance prediction of indirect evaporative coolers.