Research On Fault Location Method Of All-parallel AT Tranction System Based On BP Neural Network

With the rapid development of the high-speed railroad, the all-parallel AT has become the main power supply mode. On the one hand, the characteristics of all-parallel AT power supply mode, such as the longer power supply distance, stronger power supply capacity, more suitable for high-speed, etc. makes it satisfy the demand of the high-speed railway. On the other hand, as adopted the all-parallel AT power supply, the structure of traction system has became more complex, and the short circle fault are easier to happen. And thus a variety of researches about the theory of fault location improvement has been on the urgent. Therefore, the study of finding an effective way to solving the traction system fault location problem, is particularly important.Firstly, this thesis analysis the current distribution when the all-parallel AT traction system happened short-circuit fault. Based on this, discuss the applicability of the three traditional fault location theory, comprehensive considerate with the project economy, and detailed analysis these theories'superiority and inferiority.Second, emphasis on theoretical analysis the influence of the AT leak impedance to "AT current ratio" fault location theory. On the basis of the all-parallel AT equivalent circuit, the author derive a new "AT current ratio" fault location formula. Considering of engineering practice, the author has given a new fault location method for all-parallel AT traction system, which is basing on BP neural network theory. Training the BP neural network with the training pattern, then the equivalent impedence parameters of an unknown traction power supply system can be calculated, and those parameters can santisfy the new fault location formula which has been derived, so the BP network can calculate the position of the fault point accurately.Finally, in order to validate the correction and applicability of the BP fault location method, the thesis designe a BP network based on Delphi software, and simulates and analyzes it by using MATLAB.