Research On Io T Anomaly Detection Technology For Gas Stations Based On Deep Learning

Entering the third decade of the 21st century,science and technology have developed rapidly,and Io T devices are widely used in industry.To keep up with the times,gas stations are also transforming towards digitalization.Gas station time series data contains a lot of information,but gas data is large and complex,and there are hidden relationships between various variables.Current anomaly detection methods often produce “false anomalies” or miss real anomalies,and cannot effectively detect abnormal situations in gas data.How to efficiently and accurately detect anomalies in massive gas data has become an indispensable problem in daily operation and management of gas stations.Based on the background of intelligent industry,this thesis combines Io T,time series database and deep learning technology to study the anomaly detection of gas station Io T time series data.The main research contents are as follows:First,aiming at the problem that the external network cannot obtain the gas station data in real time,this thesis proposes a data acquisition model based on Message Queue Telemetry Transport(MQTT)protocol.The data is transmitted from the inside of the gas station to the cloud through Io T technology,thus achieving the purpose of edge-cloud collaboration and improving the security of the gas station Io T system.At the same time,by setting different topics for different serial ports of different gas stations,the source of each data can be accurately distinguished,providing an accurate and reliable data source for the anomaly detection module.Second,this thesis adopts TDengine,a domestic time series database,to store gas station time series data,and uses the streaming computing function launched by TDengine 3.0 to replace complex stream processing systems.Data subscription function is no longer needed to integrate message queue products.Based on this,the data storage module is studied and designed,which simplifies the complexity of system design and reduces operation and maintenance costs.Third,this thesis adopts an anomaly detection method for gas station Io T time series data based on deep learning Transformer model and Bi LSTM model.First,Transformer is used to perform preliminary feature extraction on the input data,and then Bi LSTM is used to perform long-distance dependency feature extraction to retain the sequential features of the data.Combining Transformer and Bi LSTM algorithms can effectively improve the accuracy of detection.Fourth,based on data acquisition,data storage,anomaly detection and anomaly alarm modules,the whole gas station Io T anomaly detection system is studied and designed.Through environment construction and testing,the feasibility and effectiveness of the system are verified.