Research On Transmission Line Fault Identification Based On Multi Source Information Fusion

Transmission lines have a wide range of construction and are mostly exposed to the external environment.The operating conditions are harsh,and they are directly affected by environmental and human factors,so they are prone to failures.Once the line trips and power failures,it will seriously affect the power transmission of the system and bring inconvenience to people’s lives.It is particularly important to identify and quickly respond to transmission line faults.On the one hand,it is necessary to quickly and accurately identify the types of faults that occur on the transmission line,and accelerate the completion of the removal of the faults.On the other hand,after the faulty line trips and powers off,the multi-source information in the power grid is collected and relevant features are extracted,combined with intelligent algorithms to complete the identification of the root cause of the fault,thereby reducing the time for the operation and maintenance personnel to check the cause of the fault on the spot,which is beneficial restore the normal operation of the system as soon as possible.For this reason,this article will conduct in-depth research on two aspects: identification of transmission line fault type and fault root cause.(1)Research on the identification method of transmission line fault type.First,PSCAD/EMTDC software is used to build a transmission line fault simulation model,and an unbalanced sample set is constructed according to the probability of occurrence of each fault type of the actual transmission line.Secondly,in view of the problem that the imbalanced data set will cause the model to identify the minority samples with poor accuracy,the BorderlineSMOTE(BSMOTE)algorithm is proposed to balance and optimize the original sample set.Then,in view of the traditional fault type identification process needs to design the fault waveform feature extraction method manually,this process is easy to cause the loss of important features,this paper proposes to use Convolutional Neural Network(CNN)with the ability to extract features and classify data as the identification model of transmission line fault types.Therefore,in the process of data processing,the multi-channel fault current waveform data is reconstructed into a two-dimensional gray image as the input of CNN model,and the corresponding fault label is output by the model,so as to complete the identification of different fault types.Finally,the experimental results show that the proposed CNN transmission line fault type identification method based on the current waveform can effectively improve the identification ability of minority samples,and can quickly and accurately complete the identification of the fault type of the transmission line,and the model has Strong anti-noise performance.(2)Research on the identification method of the root cause of the transmission line fault.First,analyze the multi-source information of field fault samples and the mechanism of transmission line faults caused by typical lightning strikes,external force damage and floating objects,etc.,and preliminary information related to the root cause of the fault is obtained.Then,the internal electrical information and external environment information of the fault are extracted.Furthermore,the correlation analysis method based on mutual information is proposed to sort the extracted fault features according to the degree of correlation with the fault root cause.The redundant features are effectively removed and the optimal feature subset is selected.Finally,this paper proposes a root cause identification method of transmission line fault based on multi-source information fusion and Support Vector Machine(SVM).In the experiment,the actual fault samples are used,and the identification models based on internal electrical characteristics,external environment characteristics and comprehensive characteristics are established respectively for analysis and comparison.The experimental results show that the identification accuracy of the model based on comprehensive features is better than that of the other two cases.When the top seven features based on mutual information correlation are used,namely,weather,reclosing condition,humidity,temperature,quarter,fault type and maximum voltage waveform ratio,the model can obtain the optimal fault root cause identification performance.