Time-lapse earthquake difference prediction based on physics-informed long short-term memory coupled with interpretability boosting

Abstract

Deep learning framework based on physical constraints and improved interpretability has revolutionized 4D seismic interpretation. This study proposes a physics-informed long short-term memory (PI-LSTM) framework integrated with interpretability enhancement techniques for high-precision time-lapse seismic difference prediction, addressing key challenges in reservoir monitoring. The model embeds the first-order velocity-stress wave equation into the LSTM gating mechanism, reducing the physical residual of North Sea field data from 62.3 kPa to 15.2 kPa—a 75.6% decrement. An interpretability enhancement module combines Shapley additive explanation value dynamic weighting with physical attention templates, reducing the seasonal fluctuation of feature importance by 38% (measured as ΔS). Key innovations include adaptive geological parameter mapping, where the physical constraint weight was automatically raised to 0.89 ± 0.04 when porosity exceeded 15%. In dual benchmark tests using Society of Exploration Geophysicists Synthetic Data and North Sea Field Surveys, PI-LSTM achieved a time-lapse prediction accuracy of 0.71-2.1 ms, equivalent to a hydrocarbon interface localization error of <3 m, outperforming commercial software by 62.9%. The framework demonstrates strong versatility across 12 reservoir types, maintaining prediction stability (coefficient of variation: <12%) under varying signal-to-noise ratios (15-40 dB). For high-pressure reservoirs (>35 MPa), the model reduced the wave equation residual to 18.6 kPa, 67.5% lower than conventional LSTMs, whereas fluid displacement volume prediction deviates by only 1.8% from well data. This work establishes a new paradigm for physics-guided 4D seismic interpretation, validated through multiscale experiments spanning from core-scale rock physics (8% error in grain contact stiffness) to field-scale reserve assessment (displacement volume R2 = 0.94).

Keywords

Physics-informed long short-term memory / Time-lapse seismic data / Interpretable machine learning / Reservoir monitoring / Wave equation constraints

Funding

This research was financially supported by Mahasarakham University; 2025 Doctoral Special Support Program Project of Chengdu Jincheng College (NO.2025JCKY(B)0018), and the Key Research Base of Humanities and Social Sciences of the Education Department of Sichuan Province, Panzhihua University, Resource-based City Development Research Center Project (NO.ZYZX-YB-2404).

References

[1] Wu YH, Pan SL, Lan HQ, et al. Dynamic reservoir monitoring using similarity analysis of passive source time-lapse seismic images: Application to waterflooding front monitoring in Shengli Oilfield, China. Petrol Sci. 2025; 22(3):1062-1079. doi: 10.1016/j.petsci.2024.12.008

[2] Ding PB, Gong F, Zhang F, Li XY. A physical model study of shale seismic responses and anisotropic inversion. Petrol Sci. 2021; 18(4):1059-1068. doi: 10.1016/j.petsci.2021.01.001

[3] Liu P, Zhao J, Li Z, Wang H. Numerical simulation of multiphase multi-physics flow in underground reservoirs: Frontiers and challenges. Capillarity. 2024; 12(3):72-79. doi: 10.46690/capi.2024.03.02

[4] Wu Y, Sicard B, Gadsden SA. Physics-informed machine learning: A comprehensive review on applications in anomaly detection and condition monitoring. Expert Syst Appl. 2024; 255:124678. doi: 10.1016/j.eswa.2024.124678

[5] Lim B, Zohren S. Time-series forecasting with deep learning: A survey. Philos Trans R Soc A Math Phys Eng Sci. 2021; 379(2194):20200209. doi: 10.1098/rsta.2020.0209

[6] Cao C, Debnath R, Alvarez RM. Physics-based machine learning for predicting Urban air pollution using decadal time series data. Environ Res Commun. 2025; 7(5):051009. doi: 10.1088/2515-7620/add795

[7] Kim S, Lee D, Lee S. Performance improvement of seismic response prediction using the LSTM-PINN hybrid method. Biomimetics (Basel). 2025; 10(8):490. doi: 10.3390/biomimetics10080490

[8] Pwavodi J, Ibrahim AU, Pwavodi PC, Al-Turjman F, Mohand-Said A. The role of artificial intelligence and IoT in prediction of earthquakes. Artif Intell Geosci. 2024; 5:100075. doi: 10.1016/j.aiig.2024.100075

[9] Xu Q, Shi Y, Bamber J, et al. Physics-Aware Machine Learning Revolutionizes Scientific Paradigm for Machine Learning and Process-Based Hydrology. [arXiv Preprint]; 2023. doi: 10.48550/arXiv.2310.05227

[10] Jain P, Coogan SC, Subramanian SG, et al. A review of machine learning applications in wildfire science and management. Environ Rev. 2020; 28(4):478-505. doi: 10.1139/er-2020-0019

[11] Marano GC, Rosso MM, Aloisio A, Cirrincione G. Generative adversarial networks review in earthquake-related engineering fields. Bull Earthquake Eng. 2024; 22(7):3511-3562. doi: 10.1007/s10518-023-01645-7

[12] Kloosterman HJ, Kelly RS, Stammeijer J, et al. Successful application of time-lapse seismic data in shell expro’s gannet fields, Central North Sea, UKCS. Petrol Geosci. 2003; 9(1):25-34. doi: 10.1144/1354-079302-513

[13] Haverl MC, Aga M, Reiso E. Integrated Workflow for Quantitative use of Time-Lapse Seismic Data in History Matching: A North Sea field Case. In: Paper Presented At: SPE Europec Featured at EAGE Conference and Exhibition; 2005. Madrid, Spain. SPE-94453. doi: 10.2118/94453-MS

[14] Dong Y, Shen Y, Guo K, et al. Advanced workflow for time-lapse seismic monitoring of CO2 storage in saline aquifers with its application in a field basin. Sci Rep. 2025; 15(1):21345. doi: 10.1038/s41598-025-09476-z

[15] Koster K, Gabriels P, Hartung M, Verbeek J, Deinum G, Staples R. Time-lapse seismic surveys in the North Sea and their business impact. Lead Edge. 2000; 19(3):286-293. doi: 10.1190/1.1438594

[16] Alizadegan H, Rashidi Malki B, Radmehr A, Karimi H, Ilani MA. Comparative study of long short-term memory (LSTM), bidirectional LSTM, and traditional machine learning approaches for energy consumption prediction. Energy Explor Exploit. 2025; 43(1):281-301. doi: 10.1177/01445987241260319

[17] Botunac I, Bosna J, Matetić M. Optimization of traditional stock market strategies using the LSTM hybrid approach. Information. 2024; 15(3):136. doi: 10.3390/info15030136

[18] Kiran Kumar V, Ramesh KV, Rakesh V. Optimizing LSTM and Bi-LSTM models for crop yield prediction and comparison of their performance with traditional machine learning techniques. Appl Intell. 2023; 53(23):28291-28309. doi: 10.1007/s10489-023-05005-5

[19] Zhao C, Zhang F, Lou W, Wang X, Yang J. A comprehensive review of advances in physics-informed neural networks and their applications in complex fluid dynamics. Phys Fluids. 2024; 36(10):101101. doi: 10.1063/5.0226562

[20] Qi S, Sarris CD. Physics-informed deep operator network for 3-d time-domain electromagnetic modeling. IEEE Trans Microw Theory Tech. 2024; 72(12):5204-5216. doi: 10.1109/TMTT.2024.3521389

[21] Liu B, Pang Y, Jiang P, et al. Physics-driven deep learning inversion for direct current resistivity survey data. IEEE Trans Geosci Remote Sens. 2023; 61:1-11. doi: 10.1109/TGRS.2023.3263842

[22] Moser P, Fenz W, Thumfart S, Ganitzer I, Giretzlehner M. Modeling of 3D blood flows with physics-informed neural networks: Comparison of network architectures. Fluids. 2023; 8(2):46. doi: 10.3390/fluids8020046

[23] Fawad M, Mondol NH. Monitoring geological storage of CO 2 using a new rock physics model. Sci Rep. 2022; 12(1):297. doi: 10.1038/s41598-021-04400-7

[24] Jo TH, Djezzar S, Barajas-Olalde C, Richards T. Utilizing 3D mechanical earth models for calibration and validation in a large-scale carbon capture and storage project in North Dakota. In: Proceedings of the 17^th Greenhouse Gas Control Technologies Conference (GHGT-17). Calgary, Canada; 2024. doi: 10.2139/ssrn.5027942

[25] El Bilali A, Brouziyne Y, Attar O, Lamane H, Hadri A, Taleb A. Physics-informed machine learning algorithms for forecasting sediment yield: An analysis of physical consistency, sensitivity, and interpretability. Environ Sci Pollut Res Int. 2024; 31(34):47237-47257.doi: 10.1007/s11356-024-34245-2

[26] Saxena N. Exact results for generalized Biot-Gassmann equations for rocks that change in pore shape and grain geometry. Geophys J Int. 2015; 203(3):1575-1586. doi: 10.1093/gji/ggv386

[27] Bemer E, Hamon Y, Adelinet M. Consistent experimental investigation of the applicability of Biot-Gassmann’s equation in carbonates. Geophysics. 2019; 84(4):WA97-WA113. doi: 10.1190/geo2018-0631.1