SOURCE:

Faculty: Engineering
Department: Electrical Engineering

CONTRIBUTORS:

Aneke, J. I.
Ezechukwu, O. A.

ABSTRACT:

Fault Identification, classification and location on Ikeja West – Benin 330kV electric power transmission lines have been achieved using wavelet transform and artificial neural networks. This work proposed an improved solution based on wavelet transform and neural network back-propagation algorithm. The simulated three-phase fault current and voltage waveforms in the power transmission-line were first pre-processed and then decomposed using wavelet multi-resolution analysis to obtain the high frequency details and low frequency approximations. The data parameters of the patterns formed with the high frequency signal components were arranged as inputs of the neural network, whose task was to indicate the occurrence of a fault on the lines. The data parameters of the patterns formed using low frequency approximations were arranged as inputs of the second neural network, whose task was to indicate the exact fault type. The new method used both low and high frequency information of the fault signal to achieve an exact location of the fault.The neural networks were trained to recognize patterns and classify data. Feed forward networks have been employed along with back propagation algorithm for each of the three phases in the fault location process. Neural network (NN) designs with varying number of hidden layers and neurons per hidden layer have been provided to validate the choice of the neural networks in each step and simulation environment for real time fault diagnosis and detection was also presented. An analysis of the learning and generalization characteristics of elements in power system was presented using Neural Network toolbox in MATLAB/SIMULINK environment. Cross-Entropies of 8.1803e-3, 7.3899e-3 and 4.2980e-3 were achieved by the successfully trained NNs for the fault identification, classification and single line-to-ground fault location purposes respectively. Also the simulation results obtained from training the NNs for the fault location purposes indicated an average percentage error in the expected fault points of 0.6500%, 0.5811%, 0.7305% and 0.6447% for single line-to-ground, double line-to-ground, double line and three phase fault locations respectively with fault resistance of 15Ω and also an average percentage error of the expected results of 0.7284%, 0.7611%, 0.6837% and 0.6721% for single line-to-ground, double line-to-ground, double line and three phase fault locations respectively with fault resistance of 70Ω. These results demonstrated that wavelet multi-resolution analysis and neural network pattern recognition approachis efficient in identifying and locating faults on transmission lines as satisfactory performance was achieved especially when compared to the conventional methods such as impedance and travelling wave methods.