Research On Fault Diagnosis Of Transmission Lines Based On Wavelet Packet Analysis And Neural Network

With the rapid development of electric power system, transmission line voltage grade is increasing, transmission line distance is growing and the scale of the power network is enlarging. These factors play an important role in the safety and stability of the power system. A quick and effective phase selector can ensure the high-voltage transmission lines to achieve automatic reclosing feature. When fault occurs, it is the premise that the relay accurately and quickly identify and select the fault phase for it can operate correctly. In view of this, it proposes a method based on wavelet packet theory to identify the type of transmission line fault and study the feasibility of transmission line fault diagnosis and protection, using the wavelet analysis and wavelet packet theory in this dissertation.The wavelet analysis and wavelet packet related theory are systematically elaborated in this thesis, including the status of identification of the transmission line fault and the development and application of wavelet packet. Fully taking the actual situation of the transmission line into account, it has an important practical significance to use the330kV transmission line simulation model built in PSCAD/EMTDC platform to generate different conditions of various transmission line short-circuit fault signal.The wavelet packet has a better time-frequency characteristics than traditional wavelet analysis in the fault signal process, because of it can extract finer and more abundant fault information, so it is introduced to identify the type of transmission line fault, adopting the characteristics based on wavelet packet energy and the neural network to identify the type of transmission line fault. First, decompose of the fault current signal with the wavelet packet and calculate the energy of the band; then train the BP network using the constructed wavelet packet energy feature as the training sample; finally, testing the BP network with the test samples to achieve the function of identification for fault type and verify the validity of the proposed method. RBF neural network is especially suitable for solving the problem of fault classification with fast learning speed and the ability to approximate any continuous function at arbitrary precision. Whether it is bale to identify fault type is investigated in the same simulation conditions.It can be found that the limitations of the BP network and RBF network restrict the accuracy of the identification of fault type in actual simulation. So the BP neural network based on quantum particle swarm optimization (QPSO) is in the introduction of transmission line fault identification. It is shown that the BP network based on quantum particle swarm is in the superiority of the transmission line fault type identification in the simulation results.