Low-Level Inter-Turn Fault Detection Algorithm for Transformer Differential Protection

Abstract

This paper presents a novel hybrid protection scheme based on Maximal Overlapped Discrete Wavelet Transform (MODWT) energy and a specialized difference function (DF) to accurately detect low-level inter-turn short-circuit faults in power transformers while maintaining high-selectivity features against transient conditions. Low-level inter-turn short-circuit faults (LIFs) in power transformers start at a low level and gradually spread to other windings. It is crucial to detect the fault in early stages and prevent further damage by disconnecting the faulty transformer immediately. A wavelet transform and difference function-based Transformer Differential Protection (TDP) algorithm is proposed in this paper. A differential protection scheme consists of two stages: feature extraction and fault detection. Maximum Overlapped Discrete Wavelet Transform (MODWT) energy and a difference function are used for feature extraction and an analytical logic is used for fault detection. It is also shown that this combination provides more reliable differential protection scheme than TDP with the wavelet transform only or TDP with a difference function (DF) alone. The method is assessed with experimental datasets collected from a laboratory-based, custom-built transformer which is specifically designed for validating the methods to detect LIFs. The method is evaluated according to a confusion matrix method with accuracy, dependability and sensitivity indices. The proposed TDP method detected all LIF cases, representing less than 2% of total windings. Therefore, the proposed hybrid algorithm represents an innovative step in applied system monitoring by providing a high-precision, software-based solution that enhances the operational reliability and resilience of existing TDP systems without requiring additional hardware.