Fault monitoring and visualization of power system under common information model

Abstract

Traditional fault monitoring methods for converter valves in high-voltage direct current (HVDC) systems lack a unified standard for converter valve data, making it difficult to exchange and integrate information between different systems, and unable to timely and accurately identify potential problems in the operation of converter valves. This article utilized the Common Information Model (CIM) to achieve standardized representation and interactive operation of converter valve monitoring data, and improved the performance of fault detection for converter valves through ensemble learning. It collected data from various entities in the converter valve (such as multi-source sensor data, electrical information, and so on) and uses CIM for standardized representation. The attributes of each entity in the converter valve can be corresponding to the CIM model. This article preprocessed the collected converter valve data to ensure the quality and availability of the data. It used the Adaboost ensemble learning model to combine Support Vector Machine (SVM), Decision Tree (DT), K-Nearest Neighbors (KNN), and Naive Bayes. This article combined the CIM with ensemble learning Adaboost. The experimental results showed that CIM-Adaboost can accurately classify six common converter valve fault types. After mixing normal and faulty samples in this article, CIM-Adaboost still maintains high stability, with an accuracy rate of 94.6% for fault classification. The combination of CIM and Adaboost can effectively improve the effectiveness of converter valve fault diagnosis.