Terminal-based method for efficient inter-turn fault localization and severity assessment in transformer windings

Early detection of Inter-Turn Fault (ITF) in power transformers is crucial for preventing extensive damage to the windings. However the existing methods based on calculation of fault factors, machine learning methods, and statistical indices require additional terminals or internal tap points, extensive training data and high computational demands, and limited applicability due to non-standardization and specific frequency band selection. To overcome these limitations, this paper presents a novel, non-invasive method for identifying the location and severity of ITFs using only external terminal measurements. Our approach leverages a non-iterative subspace identification algorithm to estimate the pole-zero-gain model from the measured Driving Point Impedance (DPI) frequency response, $Z_f$(s), of a faulty transformer. By solving a matrix equation involving gain, $K$, series capacitance ($C_s$), and ground capacitance ($C_g$), we determine the faulty capacitance, $C_f$ for each winding section. The fault location is then identified by correlating the measured and estimated DPIs, with the section exhibiting the maximum correlation being deemed the faulty disc and its corresponding $C_f$ is the faulty capacitance. This approach accurately identifies the faulty disc and determines the extent of insulation damage caused by the ITF, as validated through simulation and experimental results on various winding types. It paves the way for a more efficient and reliable diagnostic solution.