Deep learning surrogate model-based randomized maximum likelihood for large-scale reservoir automatic history matching

Automatic history matching in large-scale reservoir simulations poses significant challenges due to the complexity and uncertainty inherent in reservoir parameters. In this paper, we introduced a deep learning-based surrogate model, termed Convolution Recurrent Neural Network, for addressing these challenges. The Convolution Recurrent Neural Network leverages Convolution Neural Network and Recurrent Neural Network to extract spatial and temporal features respectively to approximate the intricate map between reservoir parameters and production data. And then, through the Randomized Maximum Likelihood method, the posterior distribution of reservoir parameters is sampled by optimizing a series of perturbed objective functions. This method offers several advantages, including its ability to handle high-dimensional data, capture complex reservoir dynamics, and efficiently calibrate uncertain parameters. Through comprehensive numerical experiments on both synthetic and real-world reservoir models, we demonstrate the efficacy of the approach in enhancing the efficiency and accuracy of automatic history matching in large-scale reservoir simulations. Document Type: Original article Cited as: Zhou, W., Fu, W., Liu, C., Zhang, K., Shen, J., Liu, P., Zhang, J., Zhang, L., Yan, X. Deep learning surrogate model-based randomized maximum likelihood for large-scale reservoir automatic history matching. Computational Energy Science, 2024, 1(1): 17-27. https://doi.org/10.46690/compes.2024.01.03 References: Aanonsen, S. I., Noevdal, G., Oliver, D. S., et al. The ensemble kalman filter in reservoir engineering-a review. SPE Journal, 2009, 14(3): 393-412. Asher, M. J., Croke, B. F., Jakeman, A. J., et al. A review of surrogate models and their application to groundwater modeling. Water Resources Research, 2015, 51(8): 5957-5973. Cancelliere, M., Verga, F., Viberti, D. Benefits and limitations of assisted history matching. Paper SPE 146278 Presented at the SPE Offshore Europe Oil and Gas Conference and Exhibition, Aberdeen, UK, 6–8 September, 2011. Chen, X., Zhang, K., Ji, Z. N., et al. Progress and challenges of integrated machine learning and traditional numerical algorithms: Taking reservoir numerical simulation as an example. mathematics, 2023, 11(21): 4418. Emerick, A. A., Reynolds, A. C. EnKF-MCMC. Paper SPE 131375 Presented at the SPE EUROPEC/EAGE Annual Conference and Exhibition, Barcelona, Spain, 14–17 June, 2010. Emerick, A. A., Reynolds, A. C. Ensemble smoother with multiple data assimilation. Computers and Geosciences. 2013, 55: 3-15. Graves, A., Mohamed, A. R., Hinton, G. Speech recognition with deep recurrent neural networks. Paper Presented at IEEE international conference on acoustics, speech and signal processing, Vancouver, BC, Canada, 26-31 May, 2013. Hamdi, H., Couckuyt, I., Sousa, M. C., et al. Gaussian Processes for history-matching: application to an unconventional gas reservoir. Computational Geosciences, 2017, 21(2): 267-287. He, K. M., Zhang, X. Y., Ren, S. Q., et al. Deep residual learning for image recognition. Paper Presented at the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, Nevada, USA, 26 June - 1 July, 2016. Hochreiter, S., Schmidhuber, J. Long short-term memory. Neural Computation, 1997, 9(8): 1735-1780. Jafarpour, B., Goyal, V. K., McLaughlin, D. B., et al. Compressed history matching: exploiting transform-domain sparsity for regularization of nonlinear dynamic data integration problems. Mathematical Geosciences, 2010, 42: 1-27. Kingma, D. P., Ba, J. Adam: A method for stochastic optimization. arXiv preprint, arXiv: 1412.6980, 2014. Kitanidis, P. K. Parameter uncertainty in estimation of spatial functions: Bayesian analysis. Water Resources Research, 1986, 22(4): 499-507. LeCun, Y., Bengio, Y., Hinton, G. Deep learning. Nature, 2015, 521(7553): 436-444. Li, X., Reynolds, A. C. A gaussian mixture model as a proposal distribution for efficient markov-chain monte carlo characterization of uncertainty in reservoir description and forecasting. SPE Journal, 2020, 25(1): 1-36. Liu, N., Oliver, D. S. Evaluation of monte carlo methods for assessing uncertainty. SPE Journal, 2003, 8(2): 188-195. Lu, P., Horne, R. N. A multiresolution approach to reservoir parameter estimation using wavelet analysis. Paper SPE 62985 Presented at SPE Annual Technical Conference and Exhibition, Dallas, Texas, 1–4 October, 2000. Ma, X. P., Zhang, K., Wang, J., et al. An efficient spatial temporal convolution recurrent neural network surrogate model for history matching. SPE Journal, 2022a, 27(2): 1160-1175. Ma, X. P., Zhang, K., Zhang, J. D., et al. A novel hybrid recurrent convolutional network for surrogate modeling of history matching and uncertainty quantification. Journal of Petroleum Science and engineering, 2022b, 210: 110109 Mo, S. X., Zhu, Y. H., Zabaras, N., et al. Deep convolutional encoder-decoder networks for uncertainty quantification of dynamic multiphase flow in heterogeneous media. Water Resources Research, 2019, 55 (1): 703-728. Oliver, D. S., He, N., Reynolds, A. C. Conditioning permeability fields to pressure data. Presented at the ECMOR V-5th European Conference on the Mathematics of Oil Recovery, 1996. Oliver, D. S., Reynolds, A. C. L, N. Inverse theory for petroleum reservoir characterization and history matching. Cambridge, UK: Cambridge University Press, 2008. Paszke, A., Gross, S., Massa, F., et al. Pytorch: An imperative style, high-performance deep learning library. Advances in Neural Information Processing Systems, 2019, 32. Remy, N., Boucher, A., Wu, J. Applied geostatistics with SGeMS: A user’s guide. Cambridge, UK: Cambridge University Press, 2009. Reynolds, A. C., He, N., Chu, L., et al. Reparameterizationtechniques for generating reservoir descriptions conditioned to variograms and well-test pressure data. SPE Journal, 1996, 1(4): 413-426. Sarma, P., Durlofsky, L. J., Aziz, K., et al. Efficient real time reservoir management using adjoint-based optimal control and model updating. Computational Geosciences, 2006, 10(1): 3-36. Tang, M., Liu, Y., Durlofsky, L. J. A deep-learning-based surrogate model for data assimilation in dynamic subsurface flow problems. Journal of Computational Physics, 2020, 413: 109456. Van Leeuwen, P. J., Evensen, G. Data assimilation and inverse methods in terms of a probabilistic formulation. Monthly weather review, 1996, 124(12): 2898-2913. Wantawin, M., Yu, W., Dachanuwattana, S., et al. An iterative response-surface methodology by use of high degree-polynomial proxy models for integrated history matching and probabilistic forecasting applied to shale gas reservoirs. SPE Journal, 2017, 22(6): 2012-2031. Xiao, C., Leeuwenburgh, O., Lin, H. X., et al. Conditioning of deep-learning surrogate models to image data with application to reservoir characterization. Knowledge-Based Systems, 2021. Yan, L., Zhou, T. An adaptive surrogate modeling based on deep neural networks for large-scale bayesian inverse problems. Communications in Computational Physics, 2020, 28 (5): 2180-2205. Yu, W., Tripoppoom, S., Sepehrnoori, K., et al. An automatic history-matching workflow for unconventional reservoirs coupling MCMC and non-intrusive EDFM methods. Paper SPE 191473 Presented at the SPE Annual Technical Conference and Exhibition, Dallas, Texas, USA, 24–26 September, 2018. Zhang, J. J., Lin, G., Li, W. X., et al. An iterative local updating ensemble smoother for estimation and uncertainty assessment of hydrologic model parameters with multimodal distributions. Water Resources Research, 2018, 54(3): 1716-1733. Zhong, Z., Sun, A. Y., Wang, Y., et al. Predicting field production rates for waterflooding using a machine learning based proxy model. Journal of Petroleum Science and Engineering, 2020, 194: 107574. Zhu, Y. H., Zabaras, N. Bayesian deep convolutional encoder decoder networks for surrogate modeling and uncertainty quantification. Journal of Computational Physics, 2018, 366: 415-447.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.