Research On Overhead Transmission Line Fault Type Identification Method Based On Fault Recording Data And Meteorological Information

As the context of the power system,transmission line plays a major role in the reliable operation of the power grid.Once the transmission line fails,it will directly affect the power transmission of the power system,and then affect the normal life of the people.Power transmission lines are an important part of power transmission and transformation systems.Compared with other power transmission and transformation equipment,power transmission lines are widely distributed and numerous.Most of them are installed in open-air environments.The working conditions are complex and relatively harsh,and they are more prone to failure.Based on intelligent algorithms,it is of great significance to carry out equipment fault identification for transmission lines.However,at present,the related research methods in this field are relatively single,and the actual situation is considered less,which is difficult to meet the actual needs.In order to improve the operation level of transmission and transformation equipment,in-depth research on transmission line fault identification is urgently needed.This paper first proposes a fault classification model based on wavelet packet decomposition and recurrent neural network based on fault record data of transmission lines.Recorded wave data,as typical data under power system monitoring,has time-series characteristics,which can fully reflect the internal changes of the equipment during the fault process,and has rich information value.The recurrent neural network is deeply oriented to the sequence form,and can fully extract the change law of each electrical quantity in the fault record with time,and tap the potential characteristics.The wavelet packet decomposition is used to process the time-frequency domain of the faulty record,which improves the network's ability to recognize the high-frequency mutations contained in the record.Furthermore,comparing and optimizing the parameters of the cyclic network and the number of wavelet packet decomposition layers improves the accuracy of the obtained model.Secondly,considering the close relationship between the open-air work of the transmission line and the external meteorological environment,consider comprehensively considering fault recording and meteorological characteristics to improve the diagnostic model.The external environment information corresponding to the line fault was extracted from the meteorological database,and the correlation data of the obtained data was analyzed to establish a K-neighbor fault classification model based on meteorological characteristics.In order to obtain a comprehensive identification model,two types of features are first fused from the data level.The meteorological features are time-sequenced and spliced with the recorded sequence to form a new fusion sequence and imported into the RNN network for training.Aiming at the situation that the newly generated sequence contains a large number of invariant terms,which results in poor model training effect,feature redundancy is used to reduce feature redundancy and improve the classification accuracy of the fusion model.Further,for the case where the fusion RNN model and meteorological KNN model have a poor classification effect on a certain type of fault,the above two models are fused at the decision level through D-S theory to reduce the uncertainty in the classification process.Finally,a comprehensive fault classification model with high performance is obtained.The results of the example show that the accuracy of the newly generated comprehensive model is higher than the classification model of two single data sources.Finally,considering the data anomalies existing in the line fault record data,a method of replacing the record data based on the line topology is proposed based on the actual situation.The principle of signal similarity is introduced,and a quantitative index of the feasibility of sequence data replacement is given.Based on the fusion classification model obtained in the previous section,some fault samples are selected for verification.The feasible interval of replacement under the above index system is obtained,which proves that the proposed replacement method is effective.Sex.Further,the proposed method is compared with the traditional method for replacing abnormal data,and is used for abnormal wave recording processing,and the processed sample data is introduced into the network for training.The results show that the line topology replacement method can maintain the training effect of the model to the greatest extent,and has certain advantages over the other methods of fault recording.