Detection and correction of distorted current transformer current using wavelet transform and artificial intelligence

Secondary current in a current transformer (CT) will become distorted if the CT becomes saturated because of a large primary current and a DC component. This distorted current may leads to a malfunction in the protective relay because it receives a smaller rms current during the period of the fault occurrence. Therefore detection and correction of the distorted current caused by the saturated CT is essential. The application of wavelet transform used to detect the occurrence time of the fault and identify all saturation periods is described. Two features of the fault current were extracted. Fuzzy-c-means was used to partition all possible currents into nine clusters according to their features. Nine multi-layer feed-forward neural networks (MFNNs) were trained individually with the corresponding smaller fault current data sets to obtain corrected secondary currents at the initial stage. Takagi-Sugeno-Kang fuzzy rules were then used to integrate all the MFNN output currents into the final corrected secondary current. The proposed method was verified by SIMUUNK/Visual C and was also implemented in a field programmable gate array (FPGA, Altera Stratix EP1S25F780C5 development board). The test results show the online applicability of the proposed method.