Transmission Line Fault Location And AC/DC Line Touching Fault Identification Based On Convolutional Neural Network

High-voltage transmission lines are long and often span areas with sparsely populated,complex terrain and harsh weather.Once a fault occurs,it is extremely difficult to find.If the fault point can be accurately found after a line failure,the line can be repaired in time,which is very important for the safety,stability and economic operation of the power system.In recent years,deep learning has be carried out exploratory research on the transmission line fault location,but the existing deep learning models for fault location are more sensitive to line parameters,and corresponding simulation models need to be specially constructed for different lines to retrain,which restricts the popularization and application of this type of method.In addition,my country’s HVDC transmission lines span long distances,and the density of AC lines and DC lines crossing is high.With the frequent occurrence of severe weather in recent years,the probability of AC/DC line touching fault has greatly increased.When AC/DC line touching fault occurs,if the AC lines are blindly overlapped or the DC system is restarted blindly,a secondary fault may occur,thereby threatening the safety of the power grid.It can be seen that the accurate identification of AC/DC line touching faults is a prerequisite for the formulation of AC and DC line fault recovery strategies.However,comparing AC/DC touching lines through low-resistance with the AC line grounded through high-resistance,the electrical signal at a terminal of the AC line has very similar fault characteristics,and it is difficult to use conventional analytical methods to identify AC/DC line touching fault.In order to solve the above problems,this paper uses convolutional neural network algorithm to carry out research work on transmission line fault location and fault identification for AC/DC line touching fault.The main contents include:(1)combines the Bergeron model of the transmission line with the convolutional neural network(CNN)algorithm to establish a CNN-based similarity evaluation model of the compensation voltage waveform along the line for the transmission line time-domain faultlocation method.This method performs phase-mode conversion on the voltage and current at both ends of the line,uses the Bergeron model of the transmission line to calculate the compensation voltage waveform along the line from both ends,and multiple sets of compensation voltage waveform training samples are trained to the CNN model.The trained CNN model is used to evaluate the similarity of the compensation voltage at each observation point along the line,and locate fault based on the highest similarity evaluation of compensation voltage waveform.Finally,the PSCAD model is constructed for fault simulation to verify the proposed method.The results show that the established waveform similarity CNN model is not sensitive to transmission line parameters.For lines with different parameters,the proposed method can effectively evaluate the waveform similarity without retraining the CNN model,and complete accurate transmission line fault location.The proposed CNN-based waveform similarity evaluation method can enhance the fault tolerance of the transmission line timedomain fault location method,thereby improving the accuracy of the fault location.2)Aiming at the problem that it is difficult to distinguish between AC/DC lines touching fault and traditional AC faults,ruse CNN to extract the effective features of the input signal,and a CNN-based method is proposed for identifying the AC/DC lines touching fault.Current signals at one end of the AC line are only used,and uses CNN to automatically extract the features in the fault signal to identify the AC/DC lines touching fault of AC lines through the classification mechanism.Finally,PSCAD is used to build simulation models of AC and DC power grids with different parameter lines to verify the identification accuracy of the proposed method;at the same time,considering the influence of different voltage levels and noise,the adaptability of the proposed method is verified.