PLS Based Fault Detection and Diagnosis For HVAC Systems

Heating, Ventilation and Air Conditioning (HVAC) systems are increasingly popular for improving indoor environments, which results in growing energy consumption of buildings. Once the malfunction of one component in such a system cannot be detected and handled in time, the faults may lead to an increase in energy consumption of the whole system. Even worse, huge economic losses and serious accidents will be caused. Data-based techniques can take full advantage of the vast amounts of available process data generated during HVAC operation. Partial least squares (PLS) is more suitable for handling nonlinear data in HVAC systems compared with traditional methods such as principal component analysis (PCA). However, traditional PLS cannot detect faults timely. Therefore, this paper proposes a method based on rapid iterations of the PLS algorithm to achieve fault detection and diagnosis. This method accelerates the algorithm’s iteration by adjusting convergence bound dynamically and uses the residual contribution rates for fault location. This method is applied to a HVAC model with thirty rooms, established using TRNSYS, to validate the effectiveness of the proposed method.