Artificial Neural Network Modeling of Cyclic Steam Injection Process in Naturally Fractured Reservoirs
- Ahmet Ersahin (Pennsylvania State University) | Sercan Gul (The University of Texas at Austin) | Turgay Ertekin (Pennsylvania State University) | Cenk Temizel (Aera Energy)
- Document ID
- Society of Petroleum Engineers
- SPE Western Regional Meeting, 23-26 April, San Jose, California, USA
- Publication Date
- Document Type
- Conference Paper
- 2019. Society of Petroleum Engineers
- artificial neural networks, cyclic steam injection, artificial intelligence, naturally fractured reservoirs, enhanced oil recovery
- 12 in the last 30 days
- 129 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 9.50|
|SPE Non-Member Price:||USD 28.00|
Introducing heat into a heavy oil reservoir is a well-known enhanced oil recovery (EOR) technique. Cyclic steam injection (CSI), also called as huff and puff, is one of the most common methods used to heat the reservoir to reduce fluid viscosity. Natural fractures provide effective connections throughout the reservoir, which make naturally fractured reservoirs great candidates for steam injection. Numerical simulators are capable of designing CSI operations in dual-porosity systems. However, the use of commercial simulator can be time-consuming when a large number of cases are examined for optimizations. Artificial neural networks (ANNs) work well to solve and classify the non-linear relationships between input and output parameters.
In this paper, one forward and two inverse ANN models are proposed to discuss the performance of CSI in naturally fractured heavy oil reservoirs and to discuss smart proxy models’ mimicking ability of the commonly used numerical models. In this study, we focus on a single injection and soaking cycle for a single well. The first stage of the production starts with initial recovery, and it continues until the oil flow rate drops down to a predetermined threshold value, then, the second stage (the injection phase) takes place. Injection phase is followed by the soaking phase. Delivery of steam creates a heat chamber around the wellbore (stimulated zone) where the reservoir temperature dramatically increases. The non-uniform temperature distribution of the reservoir ends up encountering different viscosity values of the oil.
The forward ANN model successfully forecasts both cumulative oil production profiles and viscosity changes around the wellbore. Reservoir properties, rock and fluid interaction parameters, fluid properties, and injection design parameters (such as steam quality, injection rate, injection time, etc.) were used as input features. A variety of ANN architectures were tested for the minimum testing errors.
The first inverse-looking ANN model, Inverse Model 1, was designed for determining the ideal injection design parameters for a desired cumulative production profile. Once injection design parameters were selected, the forward ANN model was run for those injection design parameters for verification purposes.
Second inverse-looking ANN model, Inverse Model 2, characterizes significant reservoir parameters including; matrix porosity, matrix permeability, fracture permeability, and fracture spacing. It can be laboriously challenging and computationally costly to obtain these parameters for some fields. This ANN model is trained with injection design parameters and resulting performance indicators. All the ANN-based models are controlled by a user-friendly graphical interface for the ease of the user.
|File Size||2 MB||Number of Pages||14|
Alvarado, V., and Manrique, E. (2010). Enhanced Oil Recovery: An Update Review. Energies. (3), 1529–1575. doi:10.3390/en3091529.
Briggs, P. J. (1989, January 1). A Simulator for the Recovery of Heavy Oil from Naturally Fractured Reservoirs Using Cyclic Steam Injection. Society of Petroleum Engineers. doi:10.2118/17954-MS.
Chang, J., and Ivory, J. (2016, June 7). Steam-Air Injection Process. Society of Petroleum Engineers. doi:10.2118/180738-MS.
Kulga, B., Artun, E., and Ertekin T. (2017). Development of a Data-Driven Forecasting Tool for Hydraulically Fractured, Horizontal Wells in Tight-Gas Sands. Computers & Geosciences. (103), 99–110, ISSN 0098-3004, https://doi.org/10.1016/j.cageo.2017.03.009.
Kulga, B., Artun, E., and Ertekin T. (2018). Characterization Of Tight-Gas Sand Reservoirs From Horizontal-Well Performance Data Using An Inverse Neural Network. Journal of Natural Gas Science and Engineering. (59), 35–46, ISSN 1875-5100, https://doi.org/10.1016/j.jngse.2018.08.017.
Salazar, A., Sanchez, N., Martinez, C.. (1987). Simulation of Field Results of Cyclic Steam/Gas Injection in the Bolivar Coast, Western Venezuela. Society of Petroleum Engineers. doi:10.2118/16732-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 Steam Stimulation (CSS) Processes. Society of Petroleum Engineers. doi:10.2118/174074-MS.
Trebolle, R. L., Chalot, J. P., and Colmenares, R. (1993). The Orinoco Heavy-Oil Belt Pilot Projects and Development Strategy. Society of Petroleum Engineers. doi:10.2118/25798-MS.