|Theme||Visible||Selectable||Appearance||Zoom Range (now: 0)|
Suarez-Rivera, Roberto (W. D. Von Gonten Laboratories) | Panse, Rohit (W. D. Von Gonten Laboratories) | Sovizi, Javad (Baker Hughes) | Dontsov, Egor (ResFrac Corporation) | LaReau, Heather (BP America Production Company, BPx Energy Inc.) | Suter, Kirke (BP America Production Company, BPx Energy Inc.) | Blose, Matthew (BP America Production Company, BPx Energy Inc.) | Hailu, Thomas (BP America Production Company, BPx Energy Inc.) | Koontz, Kyle (BP America Production Company, BPx Energy Inc.)
Abstract Predicting fracture behavior is important for well placement design and for optimizing multi-well development production. This requires the use of fracturing models that are calibrated to represent field measurements. However, because hydraulic fracture models include complex physics and uncertainties and have many variables defining these, the problem of calibrating modeling results with field responses is ill-posed. There are more model variables than can be changed than field observations to constrain these. It is always possible to find a calibrated model that reproduces the field data. However, the model is not unique and multiple matching solutions exist. The objective and scope of this work is to define a workflow for constraining these solutions and obtaining a more representative model for forecasting and optimization. We used field data from a multi-pad project in the Delaware play, with actual pump schedules, frac sequence, and time delays as used in the field, for all stages and all wells. We constructed a hydraulic fracturing model using high-confidence rock properties data and calibrated the model to field stimulation treatment data varying the two model variables with highest uncertainty: tectonic strain and average leak-off coefficient, while keeping all other model variables fixed. By reducing the number of adjusting model variables for calibration, we significantly lower the potential for over-fitting. Using an ultra-fast hydraulic fracturing simulator, we solved a global optimization problem to minimize the mismatch between the ISIPs and treatment pressures measured in the field and simulated by the model, for all the stages and all wells. This workflow helps us match the dominant ISIP trends in the field data and delivers higher confidence predictions in the regional stress. However, the uncertainty in the fracture geometry is still large. We also compared these results with traditional workflows that rely on selecting representative stages for calibration to field data. Results show that our workflow defines a better global optimum that best represents the behavior of all stages on all wells, and allows us to provide higher-confidence predictions of fracturing results for subsequent pads. We then used this higher confidence model to conduct sensitivity analysis for improving the well placement in subsequent pads and compared the results of the model predictions with the actual pad results.
Abstract The objective of this study was to perform flow simulation based-reservoir modeling on a two-well pad with a long production history and identical completion parameters in the Midland Basin. A reservoir model was built using properties generated from an established geomodel. Sensitivity analysis was performed during early history match to identify ‘heavy hitters’. Subsequent history matching was conducted with less than 10% of global error, and ranges of uncertain parameters have substantially narrowed as a result. The top 50 history-matched models are selected to predict Estimate Ultimate Recovery (EUR) followed by probabilistic analysis that shows P50 of oil EUR is within acceptable range of deterministic EUR estimates. Lateral spacing sensitivity was investigated with the best history-matched model to find the maximum volume and economic benefit by varying lateral spacing of a two-well pad. The results show that, given the current completion design, well spacing tighter than the current development practice in the area is less effective in terms of volume recovery yet economic values suggest that the optimum spacing for the area is around 150% of current development assumption for one section. The presented workflow provides a systematic approach to find the optimum lateral spacing in terms of volume and economic matrices per one section. Change in commodity price will shift optimum well spacing recommendation by suggested workflow. Similar methodology can be readily performed to evaluate spacing optimization in other acreage.
Abstract Technological advancements enable natural gas to be economically produced from ultratight shale rocks. However, due to the limited availability of long-term production data as well as the complexity of gridding, for reservoir simulation studies, in dealing with hydraulic fractures, an efficient automatic history-matching workflow in a probabilistic manner for performing history matching, production forecasting, and uncertainty quantification is highly needed. This can provide critical insights for the decision-making processes. In this study, we present an integrated history-matching workflow through coupling an innovative non-intrusive EDFM (Embedded Discrete Fracture Model) method, proxy modeling of KNN (K-Nearest Neighboring), and MCMC (Markov-chain Monte Carlo) sampling. The non-intrusive EDFM method can be applied in conjunction with any third-party reservoir simulators without the need of changing the source codes. Through non-neighboring connections, EDFM can accurately and efficiently handle hydraulic fractures, which does not require local grid refinement nearby fractures. The design of experiment is applied to perform sensitivity analysis with the purpose of identifying significant uncertain parameters. The KNN is utilized to build proxy model and its quality can be improved through multiple iterations of the workflow. The classic Metropolis-Hasting (MH) algorithm of MCMC is employed to perform sampling and predict posterior distribution of uncertain parameters. An application of the workflow to a horizontal shale-gas well from Marcellus shale is demonstrated and discussed in this study. Gas desorption effect is considered in the reservoir model. Six uncertain parameters are considered for this well including matrix porosity and permeability, fracture half-length, fracture conductivity, fracture height, and fracture water saturation. Based on 10 iterations and 250 simulation cases, 52 history-matching solutions with reasonable match results against actual gas and water production rates were identified. After history matching, we performed production forecasting for 30 years using all history-matching solutions under the constraint of constant flowing bottomhole pressure of 500 psi. Reliable P10, P50, and P90 of EUR (estimated ultimate recovery) predictions of gas recovery were determined as 11.9, 13.1, and 16.4 Bcf (billion cubic feet), respectively. In addition, the narrower posterior distributions of six uncertain parameters were quantified. The values with the highest frequency for each parameter are determined: porosity is 10.4%, permeability is 0.00034 md, fracture half-length is 450 ft, fracture conductivity is 2.85 md-ft, fracture height is 87.5 ft, and fracture water saturation is 38.8%.
El Gazar, Ashraf Lotfy (Abu Dhabi Company for Onshore Petroleum Operations Ltd. (ADCO)) | Alklih, Mohamad Yousef (Abu Dhabi Company for Onshore Petroleum Operations Ltd. (ADCO)) | Sumaidaa, Saleh A. (Abu Dhabi Company for Onshore Petroleum Operations Ltd. (ADCO)) | Al Shabibi, Tariq Ali (Abu Dhabi Company for Onshore Petroleum Operations Ltd. (ADCO))
Abstract This work illustrates field development plan and optimization studies conducted on a Middle-Eastern carbonate reservoir. The field lies in an onshore area where increasing urbanization is complicating the field development with regard to safety, accessibility, and drilling sites. The reservoir exhibits relatively fair to poor reservoir characteristics and variable oil water contacts due to faulting, suggesting the presence of 5 different reservoir compartments. A total of 10 wells had penetrated the reservoir out of which 8 wells tested oil and suggested a huge initial gas cap while 2 others penetrated water leg. Six years of early production scheme (EPS, 4 producers, 1993 to 1998) data in addition to production testing, core (2 wells), MDT (3 wells), PVT (4 wells) data were gathered in order to identify the main uncertainties and test the feasibility of the full field development. EPS indicated production decline coupled with severe increase in GOR and water cut in some wells, after which the producing wells and facilities were P&A due to safety concerns and low productivity. A number of parameters were addressed and optimized during the full field development plan. These include formation evaluation and modeling parameters based on EPS findings, the limited available data, and pressure support mechanism. Several development scenarios were constructed, consisting of various combinations of horizontal producers and injectors and considering natural depletion, WI, GI, and WAG scenarios targeting the proven reserves. The dynamic modeling suggests that an ultimate recovery of 70% can be achieved by the different injection scenarios. However, considering the complexity of the surrounding environment and the size of the prize, it is recommended that the field development would be economically viable for a period of 10 years under natural depletion, provided the most effective development strategy in terms of number, location, orientation and horizontal reach is adopted.
In the wave propagation simulation by finite difference time domain (FDTD), the perfectly matched layer (PML) is often applied to eliminate the reflection artifacts due to the truncation of the finite computational domain. In the acoustic Logging-While-Drilling (LWD) FDTD simulation, due to high impedance contrast between the drill collar and fluid in the borehole, the stability and efficiency of PML scheme is critical to simulate complicated wave modes accurately. In this paper, we compare four different PML implementations in FDTD in the acoustic LWD simulation, including splitting PML (SPML), Multi-axis PML (MPML), Non-splitting PML (NPML), and complex frequency-shifted PML (CFS-PML). The simulation indicates that NPML and CFS-PML can more efficiently absorb the guide wave reflection from the computational boundaries than SPML and MPML. For large simulation time, SPML, MPML and NPML are numerically instable. However, stability of MPML can be improved further to some extent. Among all, CFS-PML is the best choice for LWD modeling. The effects of CFS-PML parameters on the absorbing efficiency are investigated, including damping profile, frequency-shifted factor, scaling factor and PML thickness. For a typical LWD case, the best value for maximum of quadratic damping profile
Abstract This paper describes the use of artificial neural networks in exploring field development strategies in conjunction with various recovery schemes. In reservoir engineering applications, the field development process is considered to be highly challenging due to inherent time constraints, lack of sufficient data, and the presence of several degrees of freedom that need to be taken into consideration. In this paper, the respective roles of the important neural network parameters with relevance to recovery scenarios under consideration are examined. The overall objective, in these recovery scenarios, is to increase the rate of oil recovery under specified GOR and WOR constraints. The increase in the oil recovery rate can be achieved by implementing an infill-drilling program or by introducing an improved recovery technique such as water or gas injection. In both cases, one of the critical decisions is to determine the locations of the new wells that need to be drilled. This paper investigates the potential use of artificial neural networks as an effective tool in identifying the optimum well locations in field development studies for various recovery schemes. The efficiency and accuracy of an artificial neural network are controlled by various parameters that are specific to a given network topology. Two of the artificial neural network parameters that control the robustness of the entire process include the learning constant, and the number of middle layer neurons. Another issue that needs to be addressed is the hidden danger stemming from the over-training of a neural network. Over-training a neural network causes the network to memorize rather than learn. A training session that concludes with memorization rather than learning will cause the artificial neural network to generate biased answers during the prediction phase. Suggestions are made in this paper to avoid the occurrence of memorization during the training process. By developing a good understanding of the roles of these key artificial neural network parameters, one will be better equipped to successfully implement the proposed hybrid (soft and hard) computing methodologies to field development studies. P. 233