Research On Fault Line Detection Method For Non- Effectively Earthed System Based On Artificial Neural Network

When non-effectively earthed system occurs the single-phase earthed fault, non-fault phase voltage will rise. This may make the weak place of line insulation breakdown and make the fault developed into phase-to-phase grounding short-circuit fault, it will cause electric power accidents and effect the operation of the power system. So, once the non-effectively earthed system occurs the single-phase earthed fault, find the fault line and remove it quickly make a big difference. This paper based on the summary of the existing fault line selection method. This draw a conclusion that the fusion of fault characteristics is the development trend. In addition, because the neural networks has strong self-learning ability, nonlinear mapping ability and fault tolerance, so this paper puts forward a method based on the BP neural network to combine the traveling wave fault features with the transient frequency fault features to find the fault line. The main works are as follows:1. The paper give a summary of the method to find the fault line for non-effectively earthed system in domestic and foreign. The paper introduced the theoretical basis and application status of all kinds of methods. The paper also evaluate the methods and point out the advantages and disadvantages of each method. The method based on the BP neural network to combine traveling wave fault features with the transient frequency fault features is the development trend in the future.2. Analyze the initial traveling wave fault characteristics and the transient frequency fault features in theory and simulation after the non-effectively earthed system occurs single phase to earth fault. We can see that: the zero mode component of the initial transient current traveling wave greatly affected by fault initial angle, while the effect of fault distance and transition resistance has a little influence; the transient frequency greatly affected by fault distance and transition resistance, while the effect of fault initial angle is has a little influence. Therefore, combine the traveling wave fault features with the transient frequency fault features to find the fault line can realize the complementary advantages of both.3. This paper introduce the BP neural network model from the aspects of basic structure, learning and training process, design method, and introduced how to improve the disadvantages of the traditional BP neural network. Making an introduction of the neural network toolbox of MATLAB. Finally puts forward a method based on the BP neural network to combine the traveling wave fault features with the transient frequency fault features to find the fault line.4. This paper verified the method that use the BP neural network combine the traveling wave fault features with the transient frequency fault features to find the fault line by simulation. Introduced how to use the simulation software ATP, how to build a non-effectively earthed system simulation model and how to set the parameters of the components in the model. Then introduced how to initialize and train the neural network model used in this paper. Finally, verified the method which is proposed in this paper can solve the problem of fault line selection of the non-effectively earthed system.