Digital Protection of Power Transformer Based on Weighted Least Square Algorithm

The paper presents a new weighted leastsquare scheme for differential protection of power transformers using digital techniques. The differential current waveform is modelled as a sum of decaying dc, fundamental and selected higher harmonic components. The model coefficients are established by using a least-square filtering technique for the sampled data. A root mean square error criterion is established to select the proper harmonic orders, data window, weighting matrix and sampling rate for fast detection of internal faults and high restraint against energising transients in transformers. A variety of simulated waveforms is presented to highlight the important features of this algorithm. The results clearly demonstrate the efficacy and weakness of the weighted least square model.