Research And Implementation Of Power Telecontrol Transmission Anomaly Detection Algorithm Based On Multidimensional Features

In recent years,smart grids based on computer science and communication technology has received an increasing attention from scholars in the field of modern engineering.Safety and stability of the power system guarantee the orderly production activities.However,with the increase of the large-scale systems and the complexity of the network structures,it is difficult to detect the network anomalies.Moreover,the accuracy and timeliness of the data are the basic conditions for the safe and stable operation of the smart grid.Therefore,it is important to find a traffic anomaly detection method to effectively avoid the loss caused by abnormal power grids.Considering the characteristics of the information transmission traffic of the power telecontrol system network,this paper proposes a multidimensional features based anomaly detection mechanism with deep learning algorithm.An anomaly detection system is also designed and implemented.The main ideas are as follow:(1)Considering the periodic characteristics of power telecontrol system,a novel traffic anomaly detection method based on improved Seq2Seq-AD is proposed.Firstly,according to the periodic characteristics of the power traffic data,timing characteristics are extracted and treated as model inputs.Then,a corresponding timing prediction model is established based on the developed Seq2 Seq algorithm.The abnormal discriminant functions are designed for the common types of anomalies in power telecontrol systems,respectively.In this case,the abnormal traffic of power telecontrol systems is detected.In addition,for the stochastic traffic data with unpredictable patterns,a ‘3σ ’ threshold judgment principle is used to determine whether the network traffic is abnormal.(2)With the developed single traffic value method,the inner structured information of the communication packet will be ignored,and cannot detect certain hidden anomalies.Therefore,structural analysis is performed on network communication data packets,and the multi-dimensional characteristics of communication messages between the dispatch master station and the local plant stations are extracted.The bidirectional LSTM algorithm is used to classify the network traffic anomalies based on the extracted features.Finally,the abnormality detection is realized based on the structured features of the electric remote control system.(3)To meet the practical demand,the requirements and design principles of the network transmission anomaly detection system are analyzed.Then,the aforementioned anomaly detection algorithm models are deployed on the designed anomaly detection system.The results show that the designed system basically satisfies the demands of power telecontrol transmission.