摘要
针对高邮台长极距地电场数据异常检测问题,提出了一种基于Informer-LSTM方法的异常检测模型。Informer模型在长时间序列特征提取上具有优势,而LSTM模型在时间序列短期相关性上表现良好,Informer-LSTM模型结合了二者的优点,非常适用于地电场数据异常检测。选取发生在高邮地震台300 km范围以内的多次地震,并对该时间段周边地电场数据进行研究。模型评估结果显示,均方误差、平均绝对误差、方差解释率等评估指标均表现良好。异常检测结果显示,模型在地震前后均检测到较为密集的异常点,表明模型异常检测结果良好。
To address anomaly detection in long-distance geoelectric field data at Gaoyou Station,we proposed an anomaly detection model based on the Informer-LSTM framework.The Informer model has advantages in extracting the feature of long-term time series,while the LSTM network model performs well in capturing short-term temporal correlations.The Informer-LSTM model combines the advantages of both,making it particularly suitable for geoelectric field data anomaly detection.In this paper,we selected multiple earthquakes occurring within a 300￣￣km radius of the Gaoyou Seismic Station and investigated the corresponding geoelectric field data in the surrounding area during this period.Evaluation results of the model show excellent performance across metrics such as Mean Squared Error(MSE),Mean Absolute Error(MAE),and Explained Variance Score.Anomaly detection results show that the model detected relatively dense anomaly points both before and after the earthquake,indicating that it performs well in anomaly detection.
作者
张珂豪
刘庆杰
陈八
孙铭
张晴
ZHANG Kehao;LIU Qingjie;CHEN Ba;SUN Ming;ZHANG Qing(Institute of Disaster Prevention,Sanhe 065201,China)
出处
《防灾科技学院学报》
2025年第2期79-88,共10页
Journal of Institute of Disaster Prevention
基金
廊坊市科学技术研究与发展计划自筹经费项目(2024011006)。
