A novel algorithm to detect internal transformer faults

The concept of Frequency Response Analysis (FRA) has been successfully used as a diagnostic technique to detect winding deformation, core and clamping structure for power transformers. The main problem about FRA techniques is to interpret the observed evolution of the frequency response in order to identify failures. Moreover, the test is conducted off-line by measuring the input/output relationships as a function of frequency in a typical range of 2 MHz to provide a current state wave form of the transformer. In this paper an alternative simple but yet effective technique is introduced to detect the internal faults within power transformer. The technique does not call for any special measuring devices as it uses the input voltage, output voltage and the input current at the power frequency and hence online monitoring can be realised. The locus of these parameters is used to identify different mechanical transformer faults. Mechanical and short circuit faults are simulated into transformer distributed parameter model to compare the differences in the proposed locus of healthy and deformed transformer.