Data-Driven Deconvolution Using Echo-State Networks Enhances Production Data Analysis in Unconventional Reservoirs
- Yuewei Pan (Texas A&M University) | Lichi Deng (Texas A&M University) | John Lee (Texas A&M University)
- Document ID
- Society of Petroleum Engineers
- SPE Eastern Regional Meeting, 15-17 October, Charleston, West Virginia, USA
- Publication Date
- Document Type
- Conference Paper
- 2019. Society of Petroleum Engineers
- Deconvolution, Pressure Transient Analysis, Echo-State Networks, Machine Learning, Production Analysis
- 6 in the last 30 days
- 74 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 9.50|
|SPE Non-Member Price:||USD 28.00|
Horizontal wells with hydraulic fractures enable economical hydrocarbon extraction from unconventional reservoirs, and the associated transient production data is a reliable source for reservoir characterization. However, the complicated convolution of rate-pressure-time history leads to a less informative analysis of true reservoir characteristics. This paper presents a novel data-driven deconvolution approach using physics-based superposition to reconstruct constant-rate-drawdown pressure response, which are further translated into diagnostic plots for efficient production analysis.
Traditional deconvolution in pressure transient analysis is usually an ill-conditioned "inverse" process that requires systematic curve-fitting, and the deconvolution response is highly sensitive to noise. Our proposed approach uses superposition equations as training features to honor the transient physics, and further projects them into higher dimensional ‘reservoir’ space (kernel-space) for the purpose of rigorous regression. Additionally, by implementing Laplacian eigenmaps, our algorithm is relatively insensitive to noise owing to its locality-preserving character. After training, the constant-rate-drawdown pressure response is reconstructed and a diagnostic plot is generated to identify key reservoir characteristics such as flow regimes.
We first validated our approach with two synthetic cases, a horizontal well with single and multiple transverse fractures (MTFW), and the drawdown pressure response was obtained through simulation using a highly variable flow rate history. Additionally, we added artificial white Gaussian noise to the simulation output to mimic measured signals collected in the field, and we input this data into our model for deconvolution. The model-reconstructed constant-rate-drawdown pressure response was used to determine flow regimes and reservoir properties such as permeability and stimulated reservoir volume (SRV) using traditional transient testing diagnostic tools and specialized plots. The deconvolved response for each case were in alignment with the fractured-basement reservoir model proposed by Kuchuk et al. (2012), and the flow regimes identified during MTFW production followed the theory proposed by Song et al. (2011). All deconvolved responses were further validated through comparison with both simulation results and analytical solutions. Through the inherent locality-preserving character, our proposed algorithm was able to handle a moderate level of noise in addition to the variation of pressure-rate signals. We then applied the methodology to a field case, and the outcomes were satisfactory.
This study showed that our proposed methodology is a reliable diagnostic tool to interpret pressure-rate data using traditional pressure transient analysis for unconventional reservoirs. Rapid and accurate deconvolved pressure response greatly enhances the analysis of data with moderate noise and highly variable production histories, enabling engineers to recognize flow patterns and estimate reservoir properties. We demonstrated the versatility and applicability of our proposed approach with synthetic and field cases.
|File Size||1 MB||Number of Pages||17|
Antonelo, E., Camponogara, E., and Foss, B. 2017. Echo State Networks for Data-Driven Downhole Pressure Estimation in Gas-Lift Oil Wells. Neural Networks, 106-117. DOI: 10.1016/j.neunet.2016.09.009.
Belkin, M. and Niyogi, P. 2003. Laplacian Eigenmaps for Dimensionality Reduction and Data Representation. Neural Computation, 15, 1373-1396. DOI: 10.1162/089976603321780317.
Bi, R., Nasrabadi, H. 2019. Molecular simulation of the constant composition expansion experiment in shale multi-scale systems. Fluid Phase Equilibria, 495, 59-68. DOI: 10.1016/j.fluid.2019.04.026.
Deng, L. and King, M.J. 2018. Theoretical Investigation of Water Blocking in Unconventional Reservoirs Due to Spontaneous Imbibition and Water Adsorption. Proceeding of Unconventional Resources Technology Conference, Houston, Texas, 23-25 July. URTEC-2875353-MS. DOI: 10.15530/URTEC-2018-2875353.
Deng, L., Olalotiti-Lawal, F., Davani, E., . 2019. Hypervolume-Based Multiobjective Optimization for Gas Lift Systems. Proceeding of SPE Oklahoma City Oil and Gas Symposium, Oklahome City, Oklahoma, 9-10 April. SPE-195208-MS. DOI: 10.2118/195208-MS.
Han, M. and Xu, M. 2018. Laplacian Echo State Network for Multivariate Time Series Prediction. IEEE Transactions on Neural Networks and Learning Systems, 29, 238-244. DOI: 10.1109/TNNLS.2016.2574963.
Jaeger, H. and Haas, H. 2004: Harnessing nonlinearity: Prediction chaotic systems and saving energy in wireless communication. Science, 304, 78-80. DOI: 10.1126/science.1091277.
Kuchuk, F. J., and Biryukov, D. 2012. Transient pressure test interpretation from continuously and discretely fractured reservoirs. Proceedings of SPE Annual Technical Conference and Exhibition, 8-10 October, San Antonio, Texas, USA. SPE-158096-MS, DOI: 10.2118/158096-MS.
Kuchuk, F., Morton, K., and Biryukov, D. 2016. Rate-transient analysis for multistage fractured horizontal wells in conventional and un-conventional homogeneous and naturally fractured reservoirs. Proceedings of SPE Annual Technical Conference and Exhibition, 26-28 September, Dubai, UAE. SPE-181488-MS. DOI: 10.2118/181488-MS.
Levitan, M. 2003. Practical Application of Pressure-Rate Deconvolution to Analysis of Real Well Tests. Proceedings of SPE Annual Technical Conference and Exhibition, 5-8 October, Denver, Colorado, USA. SPE-84290-MS. DOI: 10.2118/84290-MS.
Liu, Y., and Horne, R. 2013. Interpreting Pressure and Flow-Rate Data From Permanent Downhole Gauges by Use of Data-Mining Approaches. SPE Journal, 18, 69-82. SPE-147298-PA. DOI: 10.2118/147298-PA.
Onur, M., and Kuchuk, F. J. 2012. A new deconvolution technique based on pressure-derivative data for pressure-transient-test interpretation. SPE Journal, 17, 307-320. SPE-134315-PA. DOI: 10.2118/134315-PA.
Pan, Y., Zhou, P., Deng, L., Lee, J. 2019. Production Analysis and Forecasting for Unconventional Reservoirs Using Laplacian Echo-State Networks. Proceedings of SPE Western Regional Meeting, San Jose, California, 23-26 April. SPE-195243-MS. DOI: 10.2118/195243-MS.
Pan, Y., Bi, R., Zhou, P., Deng L., Lee, J. 2019. An Effective Physics-Based Deep Learning Model for Enhancing Production Surveillance and Analysis in Unconventional Reservoirs. Proceedings of Unconventional Resources Technology Conference, Denver, Colorado, 22-24 July. URTEC-2019-145. DOI: 10.15530/urtec-2019-145.
Song, B. and Ehlig-Economides C A. 2011. Rate-normalized pressure analysis for determination of shale gas well performance. Proceedings of North American Unconventional Gas Conference and Exhibition, 14-16 June, The Woodlands, Texas, USA. SPE-144031-MS. DOI: 10.2118/144031-MS.
Sun, Q. and Ertekin, T. 2015. The Development of Artificial-neural-network-based Universal Proxies to Study Steam Assisted Gravity Drainage(SAGD) and Cyclic Staem Stimulation (CSS) Processes. Proceedings of SPE Western Regional Meeting, 27-30 April, Garden Grove, California, USA. SPE-174074-MS. DOI: 10.2118/174074-MS.
Sun, Q. and Ertekin, T. 2017. Structuring an artificial intelligence based decision making tool for cyclic steam stimulation processes. Journal of Petroleum Science and Engineering, 154, 564-575. DOI: 10.1016/j.petrol.2016.10.042.
von Schroeter, T., Hollaender, F., and Gringarten, A. 2001. Deconvolution of Well Test Data as A Nonlinear Total Least Squares Problem. Proceedings of SPE Annual Technical Conference and Exhibition. 30 September – 3 October, New Orleans, Louisiana, USA. SPE-71574-MS. DOI: 10.2118/71574-MS.
von Schroeter, T., Hollaender, F. and Gringarten, A. 2004. Deconvolution of Well-Test Data as a Nonlinear Total Least-Squares Problem. SPE Journal, 9, 375-390. SPE-77688-PA, DOI: 10.2118/77688-PA.
Tian, C. and Horne, R. 2019. Applying Machine Learning Techniques to Interpret Flow Rate, Pressure and Temperature Data From Permanent Downhole Gauges. SPE Reservoir Evaluation & Engineering, 386-401. SPE-174034-PA, DOI: 10.2118/174034-PA.
Yang, Q., Jin, B., Banerjee, D., and Nasrabadi, H. 2019. Direct Visualization and Molecular Simulation of Dewpoint Pressure of A Confined Fluid in Sub-10 nm Slit Pores. Fuel, 235, 1216-1223. DOI: 10.1016/j.fuel.2018.08.050.
Zhang, F., Saputra, I.W.R., Niu, G.. 2018. Upscaling Laboratory Result of Surfactant-Assisted Spontaneous Imbibition to the Field Scale through Scaling Group Analysis, Numerical Simulation, and Discrete Fracture Network Model. Proceedings of SPE Improved Oil Recovery Conference, Tulsa, Oklahoma, USA. SPE-190155-MS. DOI: 10.2118/190155-MS.
Zhang, F., Saputra, I.W.R., Parsegov, S.G.. 2019. Experimental and Numerical Studies of EOR for the Wolfcamp Formation by Surfactant Enriched Completion Fluids and Multi-Cycle Surfactant Injection. Proceedings of SPE Hydraulic Fracturing Technology Conference Woodlands, TX, USA. SPE-194325-MS. DOI: 10.2118/194325-MS.