Study And Application Development Of Fault Location In Small Current Grounding System Based On Atomic Decomposition

With the accelerating pace of modernization in our country, the demand of electricity is still growing. As a result, the distribution network has an increasingly widespread use and its structure tends to be complicated. It is self-evident that the reliable, secure and stable running of distribution network is important to industrial and agricultural production and people’s daily life. According to the related statistics from power department, the neutral non-effective grounding is very common in the low-voltage distribution network(6KV~35KV) of our country, and more than 80 percents of the grounding fault is single-phase grounding fault. Due to the failure of the constitution of short circuit in this system, the short-circuit current is very small in this system, so this system is also named “small current grounding system”. The grounding fault doesn’t only affect the normal production and life, but also does great harm to the grid itself. Therefore, when the grounding fault occurs, we need to find out the fault line and the fault position as soon as possible, then restore the normal supply of power and reduce the losses to a minimum.The traditional manual method of fault line selection and fault location is not only inefficient, but also causes harm to the grid. So it is an urgent task to study an effective algorithm of fault line selection and fault location, and develop a device of fault line selection and fault location.At the basics of fault line selection, this paper proposes a fault location method of neural network based on atomic decomposition. The MATLAB simulation model is built according to the actual information from the transformer substation, then we can obtain the raw data by changing the value of the fault position, the fault phase and the grounding resistance in the simulation experiment, after the processing, we can obtain the atomic energy decomposed by the atomic decomposition, the zero-sequence active power and the fifth harmonic component as the feature quantities of fault information. Then these feature quantities are input into the established BP neural network for fault location by data fusion, the simulation results show the feasibility of the method in theory.At the meanwhile, this paper designs a device for collecting fault signal. The high-performance TMS320F2812 chip is used as the processor in this device. The device has the ability of sampling four channels of zero-sequence current signal and one channel of zero-sequence voltage signal with its high sampling speed and accuracy. Then extracting the feature quantities from the signal by the help of the powerful data processing capability of the DSP chip, and communicating with the fault diagnosis device through the CAN bus.At last, according to the actual situation, the testing platform is built in lab to test the function of the device and the feasibility of the method. After a large number of the testing experiments, the test results show the method based on atomic decomposition is feasible.