基于Prophet-GMM的大坝监测数据异常检测算法
Anomaly Detection Algorithm of Dam Monitoring Data Based on Prophet-GMM
孙政杰 1丁勇 2李登华3
作者信息
- 1. 南京理工大学 土木工程系,江苏 南京 210094
- 2. 南京理工大学 土木工程系,江苏 南京 210094;南京水利科学研究院,江苏 南京 210094
- 3. 南京水利科学研究院,江苏 南京 210094;水利部 水库大坝安全重点实验室,江苏 南京 210094
- 折叠
摘要
大坝监测数据受环境等因素影响,往往存在异常数据,异常数据的检测对于大坝的正常运行起着不可或缺的作用,但是传统异常检测算法对于大坝监测数据往往达不到精度要求.提出了一种基于Prophet-GMM的异常检测算法,利用Prophet算法较好的拟合性能对大坝数据进行拟合,由拟合数据与实测数据求残差序列,再利用GMM算法对残差序列进行聚类,从而准确识别出异常值.结果表明:Prophet-GMM法对于不同类型的大坝监测数据都能准确识别出异常值,与传统检测算法相比,在查准率、查全率及准确率 3 个检测指标上,均有较为明显的提升.
Abstract
Due to the influence of environment and other factors,there are often abnormal data in dam monitoring data and the detection of abnormal data plays an indispensable role in the normal operation of the dam.However,the accuracy of traditional anomaly detection algo-rithms for dam monitoring data often fails to meet the requirements.In this paper,an anomaly detection algorithm based on Prophet GMM was proposed.The better fitting performance of Prophet algorithm was used to fit the dam data and the residual sequence was obtained from the fit-ting data and the actual data.Then,the residual sequence was clustered by GMM algorithm to accurately identify the abnormal value.The test results show that the method proposed in this paper can accurately identify outliers for different types of dam monitoring data.Compared with the traditional detection algorithm,it has significantly improved the detection indicators of precision,recall and accuracy.
关键词
Prophet/GMM/大坝监测数据/异常检测Key words
Prophet/GMM/dam monitoring data/abnormality detection引用本文复制引用
基金项目
国家重点研发计划项目(2022YFC3005502)
浙江省水利厅科技计划项目(RB2035)
国家自然科学基金资助项目(51979174)
国家自然科学基金联合基金资助项目(U2040221)
中央级公益性科研院所基本科研业务费专项(Y321004)
出版年
2024
国家科技期刊平台
NSTL
万方数据