基于改进Prophet-LSTM-PSO的大坝异常数据检测模型

doi:10.16265/j.cnki.issn1003-3033.2025.08.0155

摘要/Abstract

摘要： 为提升大坝监测数据的异常检测性能,提出一种基于改进Prophet-长短期记忆(LSTM)-粒子群优化(PSO)的大坝异常数据检测模型。首先,通过改进Prophet法使得异常数据点位分解得到趋势分量特征;其次,将分解得到的趋势、周期和残差分量映射到三维空间,以三维空间中近邻均值距离数据代替原始时序数据;最后,结合LSTM网络与PSO算法,设定与优化异常阈值,进而实现异常数据的精准识别。结果表明:相较于传统模型,该模型在检测效果上具有明显提升,且表现出较高的稳定性。在召回率稳定维持在95%以上的前提下,精确率与准确率均超过95%,验证了该方法的有效性与实用性。

关键词: Prophet, 长短期记忆(LSTM), 粒子群优化(PSO), 异数据常检测, 大坝监测数据

Abstract:

In order to improve the anomaly detection performance of dam monitoring data, a dam abnormal data detection method based on the improved Prophot-long short term memory-particle swarm optimization Prophet-LSTM-PSO was proposed. Firstly, by improving the Prophet method, the trend component features obtained from the decomposition of abnormal data points were clearly visible. Secondly, the decomposed trend, periodic, and residual components were represented in a three-dimensional space, where the original time series data was substituted with the mean distance of the nearest neighbors in this space. Finally, abnormal data points were identified precisely by combining the LSTM network and PSO algorithm to set and optimize anomaly thresholds. The results show that the method proposed in this paper significantly improves detection performance and exhibits high stability compared with traditional methods. Notably, while maintaining a stable recall rate exceeding 95%, both accuracy and precision surpass 95%, thereby validating the effectiveness and practicality of the proposed method.

Key words: Prophet, long short term memory (LSTM), particle swarm optimization (PSO), anomaly data detection, dam monitoring data

中图分类号:

葛大龙, 丁勇, 李登华. 基于改进Prophet-LSTM-PSO的大坝异常数据检测模型[J]. 中国安全科学学报, 2025, 35(8): 164-170.

GE Dalong, DING Yong, LI Denghua. Dam anomaly detection model based on improved Prophet-LSTM-PSO[J]. China Safety Science Journal, 2025, 35(8): 164-170.

图/表 8

图1

图2

图3

表1

图4

表2

表3

表4

参考文献 20

[1]	赵二峰, 顾冲时. 混凝土坝长效服役性态健康诊断研究述评[J]. 水力发电学报, 2021, 40(5): 22-34.
	ZHAO Erfeng, GU Chongshi. Review on health diagnosis of long-term service behaviors for concrete dams[J]. Journal of Hydroelectric Engineering, 2021, 40(5): 22-34.
[2]	黄跃文, 牛广利, 李端有, 等. 大坝安全监测智能感知与智慧管理技术研究及应用[J]. 长江科学院院报, 2021, 38(10): 180-185, 198. doi: 10.11988/ckyyb.20210589
	HUANG Yuewen, NIU Guangli, LI Duanyou, et al. Research and application of intelligent perception and intelligent management technology for dam safety monitoring[J]. Journal of Yangtze River Scientific Research Institute, 2021, 38(10): 180-185, 198. doi: 10.11988/ckyyb.20210589
[3]	王建, 伍元, 郑东健. 基于多传感器信息融合的大坝监测数据分析[J]. 武汉大学学报:工学版, 2004, 37(1):32-35.
	WANG Jian, WU Yuan, ZHENG Dongjian. Analysis of dam safety monitoring data based on multi sensor data fusion theory[J]. Engineering Journal of Wuhan University, 2004, 37(1):32-35.
[4]	赵海霞, 周少娜, 肖化. 四种判别粗大误差准则的比较与讨论[J]. 大学物理实验, 2017, 30(5): 105-108.
	ZHAO Haixia, ZHOU Shaona, XIAO Hua. The comparison and discussion of four criterions of gross-error detection[J]. College Physics Experiment, 2017, 30(5): 105-108.
[5]	尚书宏, 曹亮, 鲁伟. 基于小波变换特征提取的脱介筛故障诊断算法[J]. 中国安全科学学报, 2023, 33(增2): 233-237.
	SHANG Shuhong, CAO Liang, LU Wei. Fault diagnosis algorithm for medium draining screen based on wavelet transform feature extraction[J]. China Safety Science Journal, 2023, 33(S2): 233-237. doi: 10.16265/j.cnki.issn1003-3033.2023.S2.0020
[6]	杨兴富, 刘得潭, 杨进, 等. 基于EMD-ABOD的大坝异常监测数据识别方法研究[J]. 水电能源科学, 2024, 42(6): 162-165.
	YANG Xingfu, LIU Detan, YANG Jin, et al. Research on identification method of abnormal monitoring data of dams based on EMD-ABOD[J]. Water Resources and Power, 2024, 42(6): 162-165.
[7]	赵丽军, 邢作霞, 刘洋, 等. 基于数据分析的风电机组偏航机构异常感知方法研究[J]. 太阳能学报, 2022, 43(1):108-115.
	ZHAO Lijun, XING Zuoxia, LIU Yang, et al. Research on anomaly sensing method of yaw mechanism of wind turbine based on data analysis[J]. Acta Energiae Solaris Sinica, 2022, 43(1): 108-115.
[8]	李元梦, 李登华, 丁勇. 基于DBSCAN的大坝安全监测异常数据检测算法[J]. 水电能源科学, 2024, 42(1): 149-152.
	LI Yuanmeng, LI Denghua, DING Yong. Anomaly data detection algorithm for dam safety monitoring based on DBSCAN[J]. Water Resources and Power, 2024, 42(1): 149-152.
[9]	李川洲. 基于关联规则和机器学习的大坝监测异常数据自适应识别方法研究[D]. 重庆: 重庆交通大学, 2023.
	LI Chuanzhou. Research on adaptive identification method for abnormal data of dam monitoring based on association rules and machine learning[D]. Chongqing: Chongqing Jiaotong University, 2023.
[10]	赵丹, 沈志远, 刘晓青. 基于OCISVM的矿井通风系统在线故障诊断[J]. 中国安全科学学报, 2022, 32(10): 76-82. doi: 10.16265/j.cnki.issn1003-3033.2022.10.1766
	ZHAO Dan, SHEN Zhiyuan, LIU Xiaoqing. Online fault diagnosis of mine ventilation system based on OCISVM[J]. China Safety Science Journal, 2022, 32(10): 76-82. doi: 10.16265/j.cnki.issn1003-3033.2022.10.1766
[11]	熊敏, 江德军, 高志良, 等. 大坝监测数据多维度LSTM异常检测与恢复[J]. 电子测量技术, 2023, 46(6): 51-56.
	XIONG Min, JIANG Dejun, GAO Zhiliang, et al. Dam monitoring data multi-dimensional LSTM anomaly detection and recovery[J]. Electronic Measurement Technology, 2023, 46(6): 51-56.
[12]	张志昂, 廖光忠. 改进变分自编码器的工业时序数据异常检测[J]. 计算机工程与设计, 2024, 45(1): 17-23.
	ZHANG Zhi'ang, LIAO Guangzhong. Anomaly detection of industrial time series data based on variational autoencoder[J]. Computer Engineering and Design, 2024, 45(1): 17-23.
[13]	陆旦宏, 范文尧, 杨婷, 等. 基于生成对抗Transformer的电力负荷数据异常检测[J]. 电力工程技术, 2024, 43(1): 157-164.
	LU Danhong, FAN Wenyao, YANG Ting, et al. Anomaly detection of power load data based on Transformer and generative adversarial networks[J]. Electric Power Engineering Technology, 2024, 43(1): 157-164.
[14]	刘明群, 何鑫, 覃日升, 等. 基于改进K-means聚类k值选择算法的配网电压数据异常检测[J]. 电力科学与技术学报, 2022, 37(6): 91-99.
	LIU Mingqun, HE Xin, QIN Risheng, et al. Anomaly detection of distribution network voltage data based on improved K-means clustering k-value selection algorithm[J]. Journal of Electric Power Science and Technology, 2022, 37(6): 91-99.
[15]	孙政杰, 丁勇, 李登华. 基于Prophet-GMM的大坝监测数据异常检测算法[J]. 人民黄河, 2024, 46(3): 132-135,142.
	SUN Zhengjie, DING Yong, LI Denghua. Anomaly detection algorithm of dam monitoring data based on Prophet-GMM[J]. Yellow River, 2024, 46(3): 132-135,142.
[16]	朱江行, 邹晓松, 熊炜, 等. 基于Prophet与XGBoost混合模型的短期负荷预测[J]. 现代电力, 2021, 38(3):325-331.
	ZHU Jianghang, ZOU Xiaosong, XIONG Wei, et al. Short-term power load forecasting based on Prophet and XGBoost mixed model[J]. Modern Electric Power, 2021, 38(3):325-331.
[17]	张超, 张少飞. 基于SCADA温度数据的风电机组发电机驱动端轴承异常识别方法[J]. 轴承, 2022(6):67-73.
	ZHANG Chao, ZHANG Shaofei. Anomaly recognition method for drive end bearings of wind turbine generator based on SCADA temperature data[J]. Bearing, 2022 (6): 67-73.
[18]	陈永展, 曲建岭, 王小飞, 等. 时序记忆增强的CNN-LSTM滚动轴承故障诊断方法[J]. 噪声与振动控制, 2025, 45(1):105-111.
	CHEN Yongzhan, QU Jianling, WANG Xiaofei, et al. Rolling bearing fault diagnosis based on CNN-LSTM withtime sequential memory enhancement[J]. Noise and Vibration Control, 2025, 45(1): 105-111.
[19]	郝泽嘉, 施玉群, 成博超, 等. 基于PSO-LSTM的大坝变形组合预测模型[J]. 长江科学院院报, 2025, 42(5):208-214,222. doi: 10.11988/ckyyb.20240409
	HAO Zejia, SHI Yuqun, CHENG Bochao, et al. A combined PSO-LSTM prediction model for dam deformation[J]. Journal of Changjiang River Scientific Research Institute, 2025, 42(5):208-214,222.
[20]	李松轩, 丁勇, 李登华. 基于影响因子分解法的大坝监测数据异常检测算法[J]. 人民长江, 2023, 54(4):234-240.
	LI Songxuan, DING Yong, LI Denghua. Detection method for dam abnormal monitoring data based on influcing factor decomposition[J]. Yangtze River, 2023, 54(4): 234-240.

测点类型	监测项目	测点名称
趋势类	使用水平位移计监测坝体水平位移	EX1-1、EX2-3、EX3-7、EX4-4、EX6-2
振荡类	使用钢弦式渗压计监测左岸坝体渗透压力	OH01、OH02、OH07、OH14、OH26
周期类	使用差阻式测缝计监测衬砌混凝土应变	S3-01-2、S3-02-2、S3-04-2、S3-12-1、S3-15-2

检测方法	趋势类添加			振荡类添加			周期类添加
检测方法	大量异常	中量异常	少量异常	大量异常	中量异常	少量异常	大量异常	中量异常	少量异常
LSTM-PSO	0.973	0.965	0.988	0.944	0.959	0.951	0.965	0.957	1
DBSCAN	0.836	0.922	0.954	0.74	0.785	0.864	0.91	0.877	0.896
K-means	0.6	0.558	0.62	0.527	0.512	0.594	0.49	0.534	0.473
SVM	0.96	0.96	0.971	0.928	0.928	0.916	0.984	0.95	0.953
GMM	0.786	0.771	0.871	0.63	0.736	0.852	0.856	0.76	0.799
随机森林	0.916	0.899	0.923	0.817	0.785	0.682	0.932	0.87	0.942

检测方法	趋势类添加			振荡类添加			周期类添加
检测方法	大量异常	中量异常	少量异常	大量异常	中量异常	少量异常	大量异常	中量异常	少量异常
LSTM-PSO	1	0.993	1	0.979	0.974	0.983	0.998	0.998	1
DBSCAN	0.99	0.996	0.996	0.982	0.987	0.991	0.993	0.994	0.993
K-means	0.978	0.98	0.986	0.971	0.978	0.99	0.976	0.981	0.981
SVM	0.941	0.936	0.932	0.95	0.933	0.916	0.942	0.934	0.93
GMM	0.986	0.979	0.989	0.976	0.981	0.982	0.986	0.989	0.983
随机森林	0.994	0.987	0.995	0.981	0.984	0.991	0.995	0.992	0.994

检测方法	趋势类添加			振荡类添加			周期类添加
检测方法	大量异常	中量异常	少量异常	大量异常	中量异常	少量异常	大量异常	中量异常	少量异常
LSTM-PSO	0.989	0.996	1	0.97	0.992	0.996	1	1	1
DBSCAN	0.989	0.964	0.86	0.936	0.872	0.712	0.935	0.989	0.899
K-means	0.935	0.919	1	0.944	0.88	0.844	0.942	1	0.924
SVM	0.425	0.374	0.328	0.541	0.363	0.174	0.429	0.365	0.32
GMM	0.905	0.754	0.855	0.957	0.807	0.564	0.863	0.959	0.77
随机森林	0.932	0.805	0.937	0.837	0.817	0.775	0.973	0.894	0.903