Wavelet fuzzy expert technique for classification of power transformer transients

This paper presents a wavelet-fuzzy logic technique for classifying the different transient conditions which might arise during power transformer operation. This technique utilizes the transient current captured by the differential relay used in the transformer protection. Simulation of the different conditions including external, internal and inrush current cases are created using electromagnetic transient program EMTDC. The transient differential current arising in the differential relay is analyzed into its details and approximate waveforms using the Discrete Wavelet Transform (DWT). The maximum values of the details waveform are used as an input to a fuzzy logic algorithm for building the membership functions. The input variable is divided into three variables one for the internal fault, one for the external fault and the third for the inrush current case. Triangular and trapezoidal membership functions are used for classifying the input variables which are divided into three parts medium, low and high for each type of transient. The output variable is divided into three partitions with triangular membership function representing the external fault, the internal fault and the third represents the inrush current transient cases. Six If-Then rules are implemented to represent the different classifications of the transformer transients. Rule Consequents represent external fault for low input values, internal fault for medium input values and inrush current for high input values. The fuzzy logic Matlab toolbox is used for the simulation of the suggested membership functions and rules. The integrated proposed technique is tested using different transient conditions obtained from simulation of a real power transformer. It achieves good results for different transient classification.