基于长短期记忆神经网络的鲁04井水位异常变化分析

doi:10.12196/j.issn.1000-3274.2025.02.014

摘要/Abstract

摘要： 基于长短期记忆(LSTM)神经网络, 利用鲁04井2014—2023年的井水位、气压、降水数据及固体潮参数, 分析了2023年8月6日平原M5.5地震前井水位的异常变化及其响应机制。结果显示, ① 多变量LSTM模型对预测集2022年井水位的预测均方误差(MSE)约为0.0007, 决定系数(R²)约为0.9978, 表明模型对鲁04井水位非线性时序特征具有较好的捕捉能力; ② 模型预测偏差在震前6个月呈现增大趋势, 震前1个月加速上升, 结合潮汐因子在震前1个月升至异常上限(均值+2倍标准差), 表明含水层水文地质条件可能发生变化; ③ 发震当月, 相位差从震前正负波动的混合流动模式突变为正向峰值的垂向流主导模式, 揭示了断层破裂过程中含水系统渗流状态的调整。

关键词: LSTM神经网络, 井水位预测, 固体潮, 地震前兆

Abstract: Based on Long Short-Term Memory (LSTM) neural network, anomalous variations of water level in Lu-04 well and response mechanisms before the M5.5 Pingyuan earthquake on August 6, 2023, were analyzed using well water level, air pressure, precipitation data, and solid tide parameters (2014—2023). The results show: ① The multivariate LSTM model achieved a mean squared error (MSE) of 0.0007 and a coefficient of determination (R²) of 0.9978 when predicting the well water level in 2022, indicating that the model has a good ability to capture the nonlinear time series characteristics of the Lu-04 well water level. ② The prediction bias of the model showed an increasing trend in the six months pre-earthquake, with a sharp rise in the month before the earthquake. Concurrently, the tidal factor surged to an abnormal upper limit (mean value twice the standard deviation) in the month before the earthquake, suggesting that the hydrological geological conditions of the aquifer may have changed. ③ In the month of the earthquake, the phase difference increased significantly from a mixed flow pattern (with positive-negative fluctuations) before the earthquake to a vertical flow-dominated pattern at a positive peak, revealing adjustments in the seepage state of the aquifer system during the fault rupture process.

Key words: LSTM neural network, Well water level prediction, Solid tide, Earthquake precursors

中图分类号:

P315.7

刘凯, 陈其峰, 张军, 孙豪, 王西宝. 基于长短期记忆神经网络的鲁04井水位异常变化分析[J]. 地震, 2025, 45(2): 211-228.

LIU Kai, CHEN Qi-feng, ZHANG Jun, SUN Hao, WANG Xi-bao. Analysis of Anomalous Water Level Variations in Lu-04 Well Based on Long Short-Term Memory Neural Network[J]. EARTHQUAKE, 2025, 45(2): 211-228.

参考文献

[1] 戴丹青, 席楠. 2023年8月6日山东平原M5.5地震破裂过程快速反演[J]. 中国地震, 2023, 39(3): 689-694.
DAI Dan-qing, XI Nan. Rapid inversion of the rupture process of the M5.5 earthquake in Shandong Pingyuan on August 6, 2023[J]. Earthquake Research in China, 2023, 39(3): 689-694 (in Chinese).
[2] 张海洋. 山东平原M_S5.5地震前岩石圈磁场异常变化及发震机理[J]. 大地测量与地球动力学, 2024, 44(9): 892-898+904.
ZHANG Hai-yang. Abnormal variation of lithospheric magnetic field before the Pingyuan M_S5.5 earthquake in Shandong Province and its seismogenic mechanism[J]. Journal of Geodesy and Geodynamics, 2024, 44(9): 892-898+904 (in Chinese).
[3] 高明智, 张岑, 郝冉, 等. 平原M5.5地震前后体应变异常变化分析[J]. 大地测量与地球动力学, 2024, 44(6): 643-647.
GAO Ming-zhi, ZHANG Cen, HAO Ran, et al. Analysis of abnormal changes of volumetric strain before and after Pingyuan M5.5 earthquake[J]. Journal of Geodesy and Geodynamics, 2024, 44(6): 643-647 (in Chinese).
[4] 车用太, 鱼金子. 地震地下流体学[M]. 北京: 气象出版社, 2006.
CHE Yong-tai, YU Jin-zi. Underground fluids and earthquake[M]. Beijing: China Meteorological Press, 2006 (in Chinese).
[5] Hsieh P A, Bredehoeft J D, Farr J M, et al. Determination of aquifer transmissivity from Earth tide analysis[J]. Water Resources Research, 1987, 23: 1824-1832.
[6] 高小其, 王琼, 朱成英, 等. 7级地震前地下流体前兆异常特征及判定指标的研究[J]. 内陆地震, 2012, 26(2): 123-137.
GAO Xiao-qi, WANG Qiong, ZHU Cheng-ying, et al. Research on characteristics of underground fluid precursory anomaly and decision index before earthquake with M_S7.0[J]. Inland Earthquake, 2012, 26(2): 123-137 (in Chinese).
[7] 黄辅琼. 中国大陆地震地下水观测井对大地震的响应[D]. 北京: 中国地震局地球物理研究所, 2008.
HUANG Fu-qiong. Response of wells in groundwater monitoring network in Chinese Mainland to recent large earthquakes[D]. Beijing: Institute of Geophysics, China Earthquake Administration, 2008 (in Chinese).
[8] 付虹, 赵小艳. 汶川M_S8.0地震前云南地区显著前兆观测异常分析[J]. 地震学报, 2013, 35(4): 477-484.
FU Hong, ZHAO Xiao-yan. Analysis on remarkable precursory anormalies observed in Yunnan area before Wenchuan M_S8.0 earthquake[J]. Acta Seismologica Sinica, 2013, 35(4): 477-484 (in Chinese).
[9] 刘耀炜, 任宏微, 张磊, 等. 鲁甸6.5级地震地下流体典型异常与前兆机理分析[J]. 地震地质, 2015, 37(1): 307-318.
LIU Yao-wei, REN Hong-wei, ZHANG Lei, et al. Underground fluid anomalies and the precursor mechanisms of the Ludian M_S6.5 earthquake[J]. Seismology and Geology, 2015, 37(1): 307-318 (in Chinese).
[10] 晏锐, 田雷, 王广才, 等. 2008年汶川8.0级地震前地下流体异常回顾与统计特征分析[J]. 地球物理学报, 2018, 61(5): 1907-1921.
YAN Rui, TIAN Lei, WANG Guang-cai, et al. Review and statistically characteristic analysis of underground fluid anomalies prior to the 2008 Wenchuan M_S8.0 earthquake[J]. Chinese Journal of Geophysics, 2018, 61(5): 1907-1921 (in Chinese).
[11] 杨竹转, 邓志辉, 杨跃文, 等. 2014 年鲁甸M_S6.5地震前云南丽江地下流体的异常变化[J]. 地震地质, 2018, 40(2): 480-498.
YANG Zhu-zhuan, DENG Zhi-hui, YANG Yue-wen, et al. Study for abnormal variations of underground fluid in Lijiang area, Yunnan Province before Ludian M_S6.5 earthquake in 2014[J]. Seismology and Geology, 2018, 40(2): 480-498 (in Chinese).
[12] 缪阿丽, 李锋, 王俊, 等. 2018年安徽无为M_L4.1地震地下流体前兆异常特征[J]. 地震, 2021, 41(4): 192-202.
MIAO A-li, LI Feng, WANG Jun, et al. Typical underground fluid anomalies of the Wuwei, Anhui M_L4.1 earthquake occurred in 2018[J]. Earthquake, 2021, 41(4): 192-202 (in Chinese).
[13] 车用太, 刘成龙, 鱼金子, 等. 汶川M_S8.0地震的地下流体与宏观异常及地震预测问题的思考[J]. 地震地质, 2008, 30(4): 828-838.
CHE Yong-tai, LIU Cheng-long, YU Jin-zi, et al. Underground fluid anomaly and macro anomaly of M_S8.0 Wenchuan earthquake and opinions about earthquake prediction[J]. Seismology and Geology, 2008, 30(4): 828-838 (in Chinese).
[14]王鑫, 吴际, 刘超, 等. 基于LSTM循环神经网络的故障时间序列预测[J]. 北京航空航天大学学报, 2018, 44(4): 772-784.
WANG Xin, WU Ji, LIU Chao, et al. Exploring LSTM based recurrent neural network for failure time series prediction[J]. Journal of Beijing University of Aeronautics and Astronautics, 2018, 44(4): 772-784 (in Chinese).
[15] 闫佰忠, 孙剑, 王昕洲, 等. 基于多变量LSTM神经网络的地下水水位预测[J]. 吉林大学学报(地球科学版), 2020, 50(1): 208-216.
YAN Bai-zhong, SUN Jian, WANG Xin-zhou, et al. Multivariable LSTM neural network model for groundwater levels prediction[J]. Journal of Jilin University (Earth Science Edition), 2020, 50(1): 208-216 (in Chinese).
[16] 熊波, 李肖霖, 王宇晴, 等. 基于长短时记忆神经网络的中国地区电离层TEC预测[J]. 地球物理学报, 2022, 65(7): 2365-2377.
XIONG Bo, LI Xiao-lin, WANG Yu-qing, et al. Prediction of ionospheric TEC over China based on long and short-term memory neural network[J]. Chinese Journal of Geophysics, 2022, 65(7): 2365-2377 (in Chinese).
[17] 韩盈, 安志国, 底青云, 等. 基于循环神经网络的大地电磁信号噪声压制研究[J]. 地球物理学报, 2023, 66(10): 4317-4331.
HAN Ying, AN Zhi-guo, DI Qing-yun, et al. Research on noise suppression of magnetotelluric signal based on recurrent neural network[J]. Chinese Journal of Geophysics, 2023, 66(10): 4317-4331 (in Chinese).
[18] Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.
[19] 张昭栋, 郑香媛, 殷积涛, 等. 井水位振荡试验及其结果[J]. 地震地质, 1992, 14(2): 183-188.
ZHANG Zhao-dong, ZHENG Xiang-yuan, YIN Ji-tao, et al. Oscillation test of well level and its result[J]. Seismology and Geology, 1992, 14(2): 183-188 (in Chinese).
[20] 邵志刚, 王芃. 2008年汶川8.0级地震对地震预测研究的启示思考[J]. 地震, 2018, 38(2): 1-10.
SHAO Zhi-gang, WANG Peng. Reflections on earthquake prediction research ten years after the 2008 Wenchuan M_S8.0 earthquake[J]. Earthquake, 2018, 38(2): 1-10 (in Chinese).
[21] 廖欣, 刘春平, 石云, 等. 川06井水位固体潮效应变化初探[J]. 地震学报, 2014, 36(2): 299-305.
LIAO Xin, LIU Chun-ping, SHI Yun, et al. Changes in solid tidal effect of water level in the well Chuan06[J]. Acta Seismologica Sinica, 2014, 36(2): 299-305 (in Chinese).
[22] 晏锐, 张立, 简春林. 云南曲靖井水位潮汐动态特征分析[J]. 地震学报, 2012, 34(3): 363-373.
YAN Rui, ZHANG Li, JIAN Chun-lin. Dynamic analysis of well water level tides in Qujing well, Yunnan Province[J]. Acta Seismologica Sinica, 2012, 34(3): 363-373 (in Chinese).
[23] 孙小龙, 向阳, 李源. 深井水位对地震波、固体潮和气压的水力响应以范县井为例[J]. 地震学报, 2020, 42(6): 719-731.
SUN Xiao-long, XIANG Yang, LI Yuan. Hydraulic response of water level to seismic wave, earth tide and barometric pressure in deep well: A case study of the Fanxian well in Henan Province[J]. Acta Seismologica Sinica, 2020, 42(6): 719-731 (in Chinese).
[24] 张昭栋, 刘庆国, 耿杰. 由承压井水位动态反演水井含水层的应力变化[J]. 华南地震, 1999, 19(1): 37-42.
ZHANG Zhao-dong, LIU Qing-guo, GENG Jie. Stress variation of well aquifer inverted by water level behaviors of pressure well[J]. South China Journal of Seismology, 1999, 19(1): 37-42 (in Chinese).
[25] 晏锐. 影响井水位变化的几种因素研究[D]. 北京: 中国地震局地震预测研究所, 2008.
YAN Rui. Study of several influence factor of well water level change[D]. Beijing: Institute of Earthquake Forecasting, China Earthquake Administration, 2008 (in Chinese).
[26] 丁风和, 车用太, 刘耀炜, 等. 地震观测井地下水埋藏类型判定方法及依据[J]. 地震研究, 2022, 45(2): 284-289.
DING Feng-he, CHE Yong-tai, LIU Yao-wei, et al. The method and basis of judging the groundwater type of the seismic observational well[J]. Journal of Seismological Research, 2022, 45(2): 284-289 (in Chinese).
[27] 刘春国, 李正媛, 吕品姬, 等. 数字化地震前兆台网观测数据质量评价方法[J]. 中国地震, 2017, 33(1): 112-121.
LIU Chun-guo, LI Zheng-yuan, LÜ Pin-ji, et al. The quality evaluation method of the digital observation data from the earthquake precursory observation networks[J]. Earthquake Research in China, 2017, 33(1): 112-121 (in Chinese).
[28] 戴宗辉, 郑建常, 周连庆, 等. 基于深度学习的山东平原M_S5.5地震余震检测及发震构造研究[J]. 地球物理学报, 2025, 68(3): 925-937.
DAI Zong-hui, ZHENG Jian-chang, ZHOU Lian-qing, et al. Deep-learning-based automatic aftershock detection and seismogenic structure for the M_S5.5 Pingyuan earthquake in Shandong, China[J]. Chinese Journal of Geophysics, 2025, 68(3): 925-937 (in Chinese).
[29] 张雅茜, 戴丹青, 杨志高, 等. 2023年8月6日山东平原5.5级地震震源参数初步分析[J]. 中国地震, 2023, 39(4): 902-912.
ZHANG Ya-qian, DAI Dan-qing, YANG Zhi-gao, et al. Preliminary analysis of source parameters of the M5.5 earthquake on August 6, 2023 in Pingyuan County, Shandong Province[J]. Earthquake Research in China, 2023, 39(4): 902-912 (in Chinese).
[30] Wang C Y, Dong M L, Xue L, et al. Tidal response of groundwater in a leaky aquifer-application to Oklahoma[J]. Water Resources Research, 2018, 54(10): 8019-8033.
[31] 关兆萱, 万永革, 黄少华. 2023年山东平原M5.5地震对周围区域的应力影响[J]. 中国地震, 2024, 40(2): 378-388.
GUAN Zhao-xuan, WAN Yong-ge, HUANG Shao-hua. The stress effect of the M5.5 earthquake on the surrounding area in Shandong Pingyuan in 2023[J]. Earthquake Research in China, 2024, 40(2): 378-388 (in Chinese).
[32] 李翠芹, 郑建常, 张正帅, 等. 2023年8月6日山东平原M_S5.5地震震源参数分析[J]. 地震研究, 2025, 48(2): 334-340.
LI Cui-qin, ZHENG Jian-chang, ZHANG Zheng-shuai, et al. Analysis of seismic source parameters of the Pingyuan, Shandong M_S5.5 earthquake on August 6, 2023[J]. Journal of Seismological Research, 2025, 48(2): 334-340 (in Chinese).
[33] 曹梦涵, 薛莲. 井水位同震响应特征与机理研究进展[J]. 地震研究, 2022, 45(2): 173-186.
CAO Meng-han, XUE Lian. Advances in groundwater coseismic response characteristics and mechanisms[J]. Journal of Seismological Research, 2022, 45(2): 173-186 (in Chinese).
[34] 史浙明. 地下水位同震响应特征及机理研究[D]. 北京: 中国地质大学(北京), 2015.
SHI Zhe-ming. Characteristic and mechanism of co-seismic hydrological response induced by earthquakes[D]. Beijing: China University of Geosciences, 2015 (in Chinese).
[35] Wakita H. Water wells as possible indicators of tectonic strain[J]. Science, 1975, 189(4202): 553-555.
[36] Jónsson S, Segall P, Pedersen R, et al. Post-earthquake ground movements correlated to pore-pressure transients[J]. Nature, 2003, 424(6945): 179-183.
[37] 杨竹转, 邓志辉, 高小其, 等. 新疆乌鲁木齐04号井数字化水位同震阶变的研究[J]. 中国地震, 2010, 26(3): 329-339.
YANG Zhu-zhuan, DENG Zhi-hui, GAO Xiao-qi, et al. Study on coseismic drop steps of water level based on the digital observation from Xin-04 well, Urumqi, Xinjiang[J]. Earthquake Research in China, 2010, 26(3): 329-339 (in Chinese).
[38] 刘凯, 张辉, 张军, 等. 山东省井水位对几次大地震同震响应的比较分析[J]. 地震学报, 2019, 41(1): 69-79.
LIU Kai, ZHANG Hui, ZHANG Jun, et al. Comparative analysis on coseismic response of water level in Shandong Province to several major earthquakes[J]. Acta Seismologica Sinica, 2019, 41(1): 69-79 (in Chinese).
[39] 马瑾, 郭彦双. 失稳前断层加速协同化的实验室证据和地震实例[J]. 地震地质, 2014, 36(3): 547-561.
MA Jin, GUO Yan-shuang. Accelerated synergism prior to fault instability: Evidence from laboratory experiments and an earthquake case[J]. Seismology and Geology, 2014, 36(3): 547-561 (in Chinese).