Deep net simulator (DNS): a new insight into reservoir simulation
Shahdad Ghassemzadeh A C , Maria Gonzalez Perdomo A , Manouchehr Haghighi A and Ehsan Abbasnejad BA Australian School of Petroleum and Energy Resources, Santos Petroleum Engineering Building, University of Adelaide, SA 5005, Australia.
B Australian Institute for Machine Learning, Engineering and Maths Sciences, North Terrace, University of Adelaide, Adelaide, SA 5000, Australia.
C Corresponding author. Email: shahdad.ghassemzadeh@adelaide.edu.au
The APPEA Journal 60(1) 124-132 https://doi.org/10.1071/AJ19093
Submitted: 6 January 2020 Accepted: 31 January 2020 Published: 15 May 2020
Abstract
Reservoir simulation plays a vital role as a diagnostics tool to better understand and predict a reservoir’s behaviour. The primary purpose of running a reservoir simulation is to replicate reservoir performance under different production conditions; therefore, the development of a reliable and fast dynamic reservoir model is a priority for the industry. In each simulation, the reservoir is divided into millions of cells, with fluid and rock attributes assigned to each cell. Based on these attributes, flow equations are solved through numerical methods, resulting in an excessively long processing time. Given the recent progress in machine learning methods, this study aimed to further investigate the possibility of using deep learning in reservoir simulations. Throughout this paper, we used deep learning to build a data-driven simulator for both 1D oil and 2D gas reservoirs. In this approach, instead of solving fluid flow equations directly, a data-driven model instantly predicts the reservoir pressure using the same input data of a numerical simulator. Datasets were generated using a physics-based simulator. It was found that for the training and validation sets, the mean absolute percentage error (MAPE) was less than 15.1% and the correlation coefficient, R2, was more than 0.84 for the 1D oil reservoirs, while for the 2D gas reservoir MAPE < 0.84% and R2 ≈1. Furthermore, the sensitivity analysis results confirmed that the proposed approach has promising potential (MAPE < 5%, R2 > 0.9). The results agreed that the deep learning based, data-driven model is reasonably accurate and trustworthy when compared with physics-derived models.
Keywords: deep learning, fluid flow, machine learning, proxy model, reservoir simulation, stand-alone simulator.
Shahdad Ghassemzadeh is a PhD student in Petroleum Engineering at the University of Adelaide. His research revolves mostly around the areas of application of machine learning and data analytics in the oil and gas industry. His current research focuses on the development of a novel artificial intelligence-based simulator for conventional and unconventional gas reservoirs. His work seeks to speed up reservoir simulation to assist history matching and field development study. In addition to research, he has been a Teaching Assistant in several courses in the Australian School of Petroleum and Energy Resources. His research interests are in reservoir modelling and simulation, unconventional resources, data-driven modelling and machine learning, proxy modelling, history matching and optimisation. |
Mary Gonzalez is a senior lecturer of the Australian School of Petroleum and Energy Resrouces (ASPER) at the University of Adelaide. Her research and teaching focus is on reservoir and production engineering, particularly production enhancement and optimisation. She joined the ASP in 2009 after several years of experience in the oil and gas industry, where she provided practical petroleum engineering, consultancy services and solutions in the areas of subsurface and production engineering. Mary has published several articles in peer-reviewed journals and presented at international conferences. She has served as a reviewer for different journals and as a mentor for young professionals, and she is the Community Education Chair and the ASPER Faculty Officer for the SPE. |
Manouchehr Haghighi is an Associate Professor of Petroleum Engineering in the Australian School of Petroleum and Energy Resources at the University of Adelaide. He holds MSc and PhD degrees in Petroleum Engineering, both from the University of Southern California. His research and teaching focus is on unconventional reservoirs and reservoir simulation. He has been the main investigator of several research projects over the last 10 years at the University of Adelaide with different oil and gas companies. He has supervised 40 MSc and 15 PhD students. Manouchehr has published more than 100 articles in peer review journals and presented numerous presentations at international conferences. He has served as a reviewer for different journals, such as the Journal of Petroleum Science and Engineering, and is an active member of the SPE. |
Dr Ehsan Abbasnejad is a Research Fellow at the University of Adelaide and is a member of the Australian Institute for Machine Learning. He was awarded his PhD degree in 2015 in Computer Science from the Australian National University. His expertise is mainly in machine learning, particularly generative methods and reinforcement learning. |
References
Alenezi, F., and Mohaghegh, S. (2017). Developing a Smart Proxy for the SACROC Water-Flooding Numerical Reservoir Simulation Model. In ‘SPE Western Regional Meeting, 23–27 April, Bakersfield, California.’ (Society of Petroleum Engineers).Artun, E., Ertekin, T., Watson, R., and Al-Wadhahi, M. (2011). Development of universal proxy models for screening and optimization of cyclic pressure pulsing in naturally fractured reservoirs. Journal of Natural Gas Science and Engineering 3, 667–686.
| Development of universal proxy models for screening and optimization of cyclic pressure pulsing in naturally fractured reservoirs.Crossref | GoogleScholarGoogle Scholar |
Bangerth, W., Klie, H., Wheeler, M. F., Stoffa, P. L., and Sen, M. K. (2006). On optimization algorithms for the reservoir oil well placement problem. Computational Geosciences 10, 303–319.
| On optimization algorithms for the reservoir oil well placement problem.Crossref | GoogleScholarGoogle Scholar |
Beckner, B. L., Haugen, K. B., Maliassov, S., Dyadechko, V., and Wiegand, K. D. (2015). General Parallel Reservoir Simulation. In ‘Abu Dhabi International Petroleum Exhibition and Conference, 9–12 November, Abu Dhabi, UAE.’ (Society of Petroleum Engineers).
Cardoso, M. A., and Durlofsky, J. L. (2010). Use of Reduced-Order Modeling Procedures for Production Optimization. SPE Journal 15, 426–435.
| Use of Reduced-Order Modeling Procedures for Production Optimization.Crossref | GoogleScholarGoogle Scholar |
Ghassemzadeh, S., and Charkhi, A. H. (2016). Optimization of integrated production system using advanced proxy based models: A new approach. Journal of Natural Gas Science and Engineering 35, 89–96.
| Optimization of integrated production system using advanced proxy based models: A new approach.Crossref | GoogleScholarGoogle Scholar |
Ghassemzadeh, S., and Pourafshary, P. (2015). Development of an intelligent economic model to optimize the initiation time of gas lift operation. Journal of Petroleum Exploration and Production Technology 5, 315–320.
| Development of an intelligent economic model to optimize the initiation time of gas lift operation.Crossref | GoogleScholarGoogle Scholar |
Ghassemzadeh, S., Shafflie, M., Sarrafi, A., and Ranjbar, M. (2013). The Importance of Normalization in Predicting Dew Point Pressure by ANFIS. Petroleum Science and Technology 31, 1040–1047.
| The Importance of Normalization in Predicting Dew Point Pressure by ANFIS.Crossref | GoogleScholarGoogle Scholar |
Ghassemzadeh, S., Schaffie, M., Sarrafi, A., and Ranjbar, M. (2014). Predicting Dew Point Pressure: Using a Hybrid Intelligent Network. Petroleum Science and Technology 32, 2969–2975.
| Predicting Dew Point Pressure: Using a Hybrid Intelligent Network.Crossref | GoogleScholarGoogle Scholar |
Goodwin, N. (2015). Bridging the Gap Between Deterministic and Probabilistic Uncertainty Quantification Using Advanced Proxy Based Methods. In ‘SPE Reservoir Simulation Symposium, 23–25 February, Houston, Texas, USA.’ (Society of Petroleum Engineers).
Knudsen, B. R., and Foss, B. (2015). Designing Shale-Well Proxy Models for Field Development and Production Optimization Problems. Journal of Natural Gas Science and Engineering 27, 504–514.
| Designing Shale-Well Proxy Models for Field Development and Production Optimization Problems.Crossref | GoogleScholarGoogle Scholar |
Li, H., and Zhang, D. (2007). Probabilistic collocation method for flow in porous media: Comparisons with other stochastic methods. Water Resources Research 43, W09409.
| Probabilistic collocation method for flow in porous media: Comparisons with other stochastic methods.Crossref | GoogleScholarGoogle Scholar |
Mohaghegh, S. D., Gaskari, R., and Maysami, M. (2017). Shale Analytics: Making Production and Operational Decisions Based on Facts: A Case Study in Marcellus Shale. In ‘SPE Hydraulic Fracturing Technology Conference and Exhibition, 24–26 January, The Woodlands, Texas, USA.’ (Society of Petroleum Engineers).
Sbiga, H. M., and Potter, D. K. (2017). Prediction of Resistivity Index by Use of Neural Networks With Different Combinations of Wireline Logs and Minimal Core Data. SPE Reservoir Evaluation & Engineering 20, 240–250.
| Prediction of Resistivity Index by Use of Neural Networks With Different Combinations of Wireline Logs and Minimal Core Data.Crossref | GoogleScholarGoogle Scholar |
Schmidhuber, J. (2015). Deep learning in neural networks: an overview. Neural Networks 61, 85–117.
| Deep learning in neural networks: an overview.Crossref | GoogleScholarGoogle Scholar | 25462637PubMed |
Shams, R., Esmaili, S., Rashid, S., and Suleymani, M. (2015). An intelligent modeling approach for prediction of thermal conductivity of CO2. Journal of Natural Gas Science and Engineering 27, 138–150.
| An intelligent modeling approach for prediction of thermal conductivity of CO2.Crossref | GoogleScholarGoogle Scholar |
van Doren, J. F. M., Markovinović, R., and Jansen, J.-D. (2006). Reduced-order optimal control of water flooding using proper orthogonal decomposition. Computational Geosciences 10, 137–158.
| Reduced-order optimal control of water flooding using proper orthogonal decomposition.Crossref | GoogleScholarGoogle Scholar |
Wood, D. A. (2014). Optimization applications seeking to improve performance of natural gas reservoirs, production, processing and utilization: A selection of case studies. Journal of Natural Gas Science and Engineering 21, A1–A3.
| Optimization applications seeking to improve performance of natural gas reservoirs, production, processing and utilization: A selection of case studies.Crossref | GoogleScholarGoogle Scholar |
