Research On Integrated Fault Line Detection In Non-Effectively Grounded Neutral System

The neutral non-effectively grounded system (small current grounded system) is used in most of China's medium voltage distribution networks to improve the system reliability. However, when a single-phase-to-earth fault occurs, the problem of line detection is still very difficult to solve, because the low accuracy of the existing detection theoretics in use can not meet the requirements. Therefore, a further study on line detection for single-phase-to-earth fault is important and significant.In this paper, the system operation characteristics in both neutral isolated and neutral petersen-coil-grounded modes are analyzed, and various fault features when the single-phase-to-earth fault occurs are discussed. The analysis show that the line detection based on a single fault feature can't be applicable in all cases, so the integrated one based on multiple features is given.Three complementary fault characteristics, containing transient component, harmonic components and active component of zero-sequence current, are chosen as the integrated line detection basis. And three existing line detection methods based on each characteristic are listed and explained, which are wavelet packet method, fifth harmonic method and active component method. To establish a decision table, four components, the transient one, fifth harmonic one, the active one and the steady one, and the fault criterion derived from these detection methods are took as the condition attributes and the line fault estimation as the decision-making attributes. And then on the basis of the most simple decision rule derived from the table simplified, the integrated line detection method based on rough set is obtained.The integrated line detection based on rough set has a high accuracy but a week anti-jamming ability. Therefore, the line detection method based on RSNN is given, which uses the original fault data after rough set pretreatment as the inputs of the neural networks and the line detection criterion as the output. The validation results show that this method effectively improves the detection reliability. A model of neutral non-effectively grounded system is established using Matlab. By changing grounding resistance, switch time, grounding location, active component and fifth harmonic component content of zero-sequence current, various situations of the single-phase-to-earth fault are simulated, and as well as situations of arc-grounding using intermittent short circuit of grounding point. The experimental results show that the integrated line detection method based on neural networks work well in various situations.