Mining Method of SER Fault Event Set in Converter Station Based on Improved Association Rule Algorithm

As the complexity and intelligence of converter station equipment increase, on-site operation and maintenance personnel face challenges in managing the vast number of Sequential Event Recording (SER) events generated by the system in a short time frame. To address this issue, a mining method for SER fault event sets based on an improved association rule algorithm is proposed. First, data preprocessing is performed using DBSCAN clustering and the entropy method to reduce the number of events, decrease the volume of SER data, and enhance the correlations between events. Next, an item coding method is employed to refine the Apriori algorithm, allowing for the mining of support groups for SER-associated events and identifying strongly associated SER events within the converter station. Finally, the feasibility of the proposed method is demonstrated through a case study utilizing trial operation data from a converter station in a flexible DC project. The results indicate that this data mining approach effectively extracts fault feature information from a large set of SER events, thereby assisting in fault diagnosis.