Health status assessment of chillers based on grey clustering and SVDD

Aiming at the problem that data acquisition is limited and it is difficult to collect a large number of the chiller’s samples in actual industrial environment, a health status assessment method combining grey clustering and support vector data description (SVDD) is proposed. First, during the normal operation of the chiller, the SVDD model is constructed based on normal state samples. Then, fuzzy theory is used to map the relative Euclidean distance from the fault samples to the SVDD center into a health index. Finally, the grey clustering method is used to classify the health indexes into levels, thereby achieving an accurate description of the health status of the chiller. The proposed method is validated using the ASHRAE RP-1043 dataset and the actual operating data from a building’s chillers. The results show that this method can effectively assess the health status of the chiller under the condition of limited samples, and the assessment results are consistent with the actual health status of the chiller.