Fault Detection of HVDC Transformer Windings using Impedance Protection

Abstract

Power transformers play a key role in high voltage direct current (HVDC) systems to overcome the limitations of conventional AC transmission and are expensive and fundamental components in power systems. Transformer winding faults are among the most frequent issues in transformers. Since winding faults are inherently aggravated, transformer winding short circuits need to be detected at an early stage. The existing transformer protection methods encounter difficulties in detecting winding faults in transformers. This study investigates and develops a reliable method for transformer winding fault detection based on impedance protection by calculating the winding impedance utilizing the terminal voltage and current measurements, and comparing the impedance values under fault conditions with their steady-state condition values. A three-phase transformer was modelled in PSCAD simulation software using three multi-winding single-phase transformers to model internal faults. Turn-to-turn faults were simulated for different fault locations, and analyzed impedance values of each scenario using a conventional impedance protection method and Machine Learning algorithms (ML) i.e. Support Vector Machine (SVM), Decision Tree and Artificial Neural Networks (ANNS) to detect and classify the winding faults and identify their locations. The fault studies were conducted on a symmetric monopolar Voltage Source Converter (VSC) based HVDC system. The terminal voltage and current measurements were utilized to derive impedance values for each fault condition. The obtained voltage, current, and impedance measurements were fed to train the developed Machine Learning algorithm Models. A higher accuracy is obtained by optimizing the ML model parameters in fault detection, classification, and fault location detection.