An Enhancement on Fault Detection for Imbalanced Data in Industrial Chiller Systems

In many industrial control settings of smart manufacturing industries, Heating, Ventilation, and Air Conditioning (HVAC) systems are essential for maintaining a stable environment that meets production requirements. Among these systems, chillers consume the most energy, making energy efficiency management of chiller units a crucial research topic to date. However, research in this area often faces a shortage of sample data for anomalies, leading to an imbalance in the number of samples between normal and abnormal data in modeling fault detection. This imbalance can result in biased judgments by classifiers. Therefore, this paper will compare the use of oversampling combined with undersampling to address imbalanced data with the use of Generative Adversarial Network (GAN) to generate realistic abnormal data, aiming to mitigate the issue of imbalanced data. The experimental results demonstrate that compared to oversampling and undersampling, GAN can effectively generate diverse abnormal data, enabling classifiers to achieve excellent modeling results.