A robust approach to transformer fault diagnosis: integrating time-current loci analysis with statistical feature extraction

Abstract

The accurate identification of inrush currents and inter-turn faults in power transformers is crucial for maintaining the stability and safety of electrical power systems. Traditional methods often struggle to distinguish between these phenomena due to their overlapping characteristics.
This paper introduces an innovative approach utilizing the time-current loci (TIL) method to effectively differentiate between normal operating conditions, inrush currents, and inter-turn faults. By plotting time against current over one cycle, distinct loci patterns emerge for each condition, offering a robust visual and analytical basis for classification. The methodology involves detailed extraction and analysis of key statistical features from the plotted loci.
Initially, the rate of change in time and current are derived using which the orientation of the
TIL curve is analyzed. Features like the mean of orientation and skewness of orientation are computed to capture the unique geometric properties of each condition's loci thereby, enhancing classification accuracy. Experimental validation was conducted using a series of controlled tests on 1kVA, 230V / 230 V transformer under different operating conditions. The results indicated that the time-current loci method could reliably distinguish between normal conditions, inrush currents, and interturn faults. The proposed method offers several advantages over conventional techniques. It eliminates the need for complex machine learning algorithms, relying instead on straightforward geometric and statistical analysis, making it easier to implement in real-time systems. Moreover, the visual nature of the loci plots provides an intuitive understanding of transformer behavior, aiding operators in quick decision-making.