Research On Anomaly Detection Of Environmental Monitoring Sensor Data Based On Edge Computing

In recent years,China’s big data,cloud computing,Internet of Things technology and industry have developed rapidly,and wireless sensor networks have been more widely used in the field of environmental monitoring.For the problem of sensor data anomaly detection,traditional algorithms usually only focus on the time continuity of single source data,while ignoring the space-time correlation between multi-source data,which reduces the detection accuracy to a certain extent.In addition,due to the large scale of sensor data,the centralized cloud computing platform cannot meet the requirements of real-time detection of abnormal data.Therefore,it is particularly important to design an efficient detection model and algorithm.The main work of this paper is as follows:Firstly,an anomaly detection algorithm for multi-source heterogeneous data based on time correlation is proposed for the time series of multi-source heterogeneous data.This algorithm avoids the error of anomaly detection results caused by one-sided estimation of single data by calculating the outlier distance of adjacent time series according to the distance similarity measurement method based on the introduction of sliding window technology.Then the anomaly score of each point is obtained,and the anomaly points are determined according to the threshold value as candidate anomaly points.Secondly,considering the spatio-temporal correlation of the sensing data,the temporal correlation based anomaly detection algorithm is further improved,and the spatial correlation based anomaly detection algorithm is proposed.The outliers detected based on the time series are replaced by the average value,and the spatial correlation is used to further judge the abnormal nodes.By analyzing the similarity of multi-source data in different neighborhoods,the similarity matrix is obtained,Then the confidence level of the similarity matrix sequence is judged to determine whether the abnormal data detected by the time series is the real abnormal data,which further improves the accuracy of anomaly detection.Finally,the specific implementation method of spatiotemporal association anomaly detection model is designed,and a hierarchical edge computing multi-source multidimensional data anomaly detection model composed of cloud servers,edge nodes and sensor nodes is built.Then,the implementation of spatiotemporal association anomaly detection algorithm is migrated to the edge node computing model,and edge computing distributed feature mode on computing storage and network is used,The traditional cloud platform anomaly detection task is assigned to the edge node close to the data source to perform data processing,which improves the overall efficiency of data processing and achieves the effect of load balancing and low latency data processing on the sensor side and the cloud side.Through many experiments on the data set collected from the real agricultural greenhouse,the results show that the accuracy of the algorithm for detecting abnormal data is 94.25%,and the false alarm rate is 0.97%.Compared with the traditional methods,the calculation efficiency and detection accuracy have been improved.