Power Transformer Fault Classification Based On Cumulative Distribution Transformation Utilizing Singular Value Decomposition

Abstract

Several fault classification-based protection methods may occasionally malfunction due to a variety of undesirable phenomena that occur in the transformer. It is crucial to differentiate between internal faults and external abnormal conditions in order to safeguard a power transformer in its entirety. The only approach to ensure unit transformer protection is through the detection of faults within power transformers. Existing relays malfunction under exceptional circumstances such as magnetizing inrush, current transformer (CT) saturation, and high resistance internal fault conditions. Therefore, it is crucial to distinguish between the internal fault and the external abnormality or fault in the scheme of transformer protection. In this study, a novel cumulative distribution transformation (CDT)-based scheme for classifying internal power transformer faults is presented. Power transformers, the indirect symmetrical phase angle regulators (ISPAR) series, and ISPAR exciting units are used to create internal faults. 13 different categories of faults are classified after the processing and analysis of 88,128 internal fault cases. Singular value decomposition (SVD) is used after CDT dataset extraction and cross-validation, and faults are identified employing a confusion matrix. The procedure was performed utilizing Matlab® 2018a, which resulted in a 96.66% overall accuracy with reduced computing time and noise reduction.