Research And Application Of Seismic Precursor Anomaly Monitoring Technology Based On Electromagnetism Data Analysis

Earthquakes are natural disasters that is harmful for the people and society.Violent earthquakes in history have caused huge damage to people’s lives and properties.Studies have found that some physical quantities may change abnormally before occurrence of earthquakes.These pre-earthquake anomalies are called seismic precursors.Study of earthquake precursors is of great significance for understanding and predicting earthquakes.Electromagnetic precursor is one of the known precursors which is sensitive and easy-to-observe,among which the geomagnetic field and the ionospheric total electron content(TEC)have shown a certain seismic correlation in case analysis and statistical analysis of many earthquakes.This paper studies seismic precursory anomaly detection algorithm using these two kinds of electromagnetic data.Firstly,the correlation of geomagnetic data and earthquake is analyzed,and a seismic precursor anomaly detection method based on graph neural network is given.Short-term gaps in data are filled based on the characteristics of the original data.Then,in order to solve data missing problem and to conduct multi-station analysis,a seismic precursor anomaly detection method based on graph neural network is proposed.A mapping layer is set up to normalize the data feature from different stations,a random vertex drop layer is used to simulate missing data,and the station attention weight layer is introduced to model the weights of different stations.The model detected anomalies before 85.14%earthquakes,and the precision reached 67.52%.Next,the correlation between TEC anomalies and earthquake precursors is analyzed,and a detection method for TEC data is proposed.Firstly,the raw data is displayed and analyzed,and the characteristics of periodicity and inhomogeneity are summarized.Then,a spatiotemporal anomaly detection method is designed based on the above characteristics,and it is found that the TEC above epicenter before earthquakes will have obvious anomalies.Finally,two methods are designed to detect clustering and persistent anomalies,one method is based on unsupervised clustering and the other is based on anomaly score evaluation.The results show that both methods can detect TEC precursor anomalies well.Considering that seismic precursors can be found in both kinds of electromagnetic data,precursors before two earthquakes are analyzed,and the features of the two are combined to detect seismic precursors.First of all,it is found that precursor anomalies of geomagnetic and TEC data often occur in different days,indicating that earthquake precursor signals in the two are different,and joint analysis is of practical significance.Later,an electromagnetic data fusion framework is given,which extracts geomagnetic and TEC features and put them into a fully connected network for classification after concatenating.Experimental result shows that in the task of precursor detection before earthquakes of which magnitude is greater than 5,fused feature performs better than single feature.Finally,a cloud platform for seismic electromagnetic data acquisition and analysis is designed and developed.Firstly,the overall requirements of the platform are analyzed,and the framework is designed.Then,based on the actual needs of data acquisition function,station monitoring function and seismic precursor monitoring function,a data acquisition module with strong scalability and maintainability is developed,a low-latency and out-of-the-box video monitoring solution is designed,and a loosely coupled,a highly flexible seismic precursor monitoring algorithm integration solution is used.The system can facilitate the daily work and is helpful for analysis and research,which has certain practical value.