A New Method Of Neural Network Fault Location For UHV DC Transmission Line

With the characteristics of transferring large amounts of power over long distance more economically, interconnecting asynchronous systems and having narrow line corridor, the UHVDC transmission system has its own advantage in transferring large amounts of power over long distance. Accurately locating faults in long UHVDC transmission lines is very important in corrective and preventive fault conditions to maintain continuity of power transmission. Most of the methods proposed so far for fault locating in HVDC transmission lines are based on the traveling-wave theory. Generally, the traveling-wave based fault location methods can be classified as the single-ended and double-ended types. The methods based on double-ended measurements are more accurate; however, it is strict to the sampling devices and needs synchronizing measurements of both ends.This paper bases its idea on analyzing the transmission characteristics of traveling wave on UHVDC transmission line. Some fixed nonlinear relationship, which don’t change with the fault relationship were found between the fault distances and fault characteristic quantities. BP neural networks were adopted to fit those relationships, and then some fault location methods for UHVDC transmission line using BP neural network based on the characteristic quantities were proposed. The specific work of this paper is as follows:(1) A novel two terminal fault location method used ANN for UHVDC transmission line based on the attenuation characteristic of signals with high frequency is proposed. The precision of traditional fault location method for UHVDC transmission line based on the attenuation characteristic of signals with high frequency depends on the accuracy of attenuation constant of transmission line, but it is difficult to compute that accurately, ANN that is a good tool for non-linear data modeling and non-linear function curve-fitting is selected to solve the aforementioned problems. A variety of transmission line fault situation are simulated based on the PSCAD, and this proposed two terminal fault location method used ANN for UHVDC transmission line based on the attenuation characteristic of signals with high frequency shows satisfactory performance.(2) We calculated the ratio between the first wavelet transform modulus maximum (WTMM) of the aerial and ground mode voltages when fault occurs. A fixed nonlinear relationship, which does not change with the fault resistances, was found between the fault distance and the ratio. To extract the fault distance parameter, artificial neural network (ANN) is selected to fit this relationship. Simulation results show the locating precision of the proposed single-ended fault location method is within ±0.2% of the total line distance.(3) There is nonlinear relationship between fault distance and the propagation time gap between zero mode and aerial mode, BP neural network that is a good tool for non-linear function curve-fitting is selected to fit this relationship and the goal of fault location can be realized. A variety of transmission line fault situation are simulated based on the PSCAD, and this proposed single-ended fault location method used ANN for UHVDC transmission line based on propagation time gap between zero mode and aerial mode shows satisfactory performance.