The Method of Fault Line Selection Protection for Distribution Network Based on Wavelet Packet Decomposition and Multi Information Fusion
In order to solve the problem of difficult line selection caused by the frequent occurrence of single-phase ground faults in mixed lines of distribution network,a new protection method is proposed.The method uses db10 wavelet packet to decompose the zero sequence current of each outgoing line in the fifth layer to obtain the wavelet low frequency reconstruction coefficient of zero sequence current,and uses Hausdorff distance algorithm to calculate the low frequency reconstruction coefficient of each outgoing line to obtain the characteristic value of mismatch degree.Combined with the energy characteristics of wavelet packets,the normalized comprehensive wavelet energy eigenvalues of each outgoing line are obtained,and the two fault eigenvalues are taken as the input values.Combined with random forest algorithm,which has the characteristics of data fusion and does not need to set a threshold,the fault line selection discrimination model of distribution network is established.The model is trained with 256 groups of training sample data obtained under different working conditions to obtain the optimal parameters,and then verified with the remaining 48 groups of different test data,the results show that the method has high accuracy of fault discrimination under different fault conditions,and has strong applicability.At the same time,the proposed method is compared with the methods BP neural network and ELM information fusion line selection.The corresponding results show that the proposed fault line selection method has significant advantages both in classification accuracy and convergence time when two types of Gaussian white noise with the signal to noise ratio of 40 dB and 20 dB are added.
fault line selectiondb10 wavelet packetHausdorff distance algorithmzero-sequence current deviation matrixrandom forest algorithm
万方数据
维普
