Model Calibration and Performance Optimization Using Multiple-Point Geostatistics and Machine Learning Techniques
| dc.contributor.advisor | Chen, Zhangxing (John) | |
| dc.contributor.advisor | Costa Sousa, Mario | |
| dc.contributor.author | Khani, Hojjat | |
| dc.contributor.committeemember | Nghiem, Long | |
| dc.contributor.committeemember | Tahmasebi, Pejman | |
| dc.contributor.committeemember | Maurer, Frank | |
| dc.contributor.committeemember | Mackay, Eric J. | |
| dc.date | Fall Convocation | |
| dc.date.accessioned | 2022-11-15T17:43:09Z | |
| dc.date.embargolift | 2022-09-14 | |
| dc.date.issued | 2020-09-14 | |
| dc.description.abstract | The overall objective of reservoir modeling is to reduce the uncertainty of production forecasts by including all available data into the model. Most importantly, the model must be able to match the historical production while preserving geological data. In this work, some new techniques are developed to model the impact of geological uncertainty on history matching and optimization problems. Most of traditional history matching and optimization methods are based on some forms of trial and error schemes. These methods are not usually designed to include and/or preserve the geological continuity of the facies. The Probability Perturbation Method (PPM) is a pixel-based data integration method, which employs the power of multiple-point geostatistics (MPS), to match production data while being constrained to other reservoir data including prior geological information. Using the PPM, one can systematically perturb a spatial distribution of reservoir properties while maintaining the pattern and consistency of the geological information. One of the shortcomings of the original PPM is that there is no straightforward approach to intelligently explore the space of uncertainty (the variability between realizations). In this work, a cluster-aided probability perturbation method (CAPPM) is developed, which can efficiently search the space of uncertainty by clustering the generated realizations. This method allows us to navigate through realizations and approximate the uncertainty bounds with fewer iterations and, consequently, less computational cost. In addition, a new segmentation method, which does not require flow simulation, is presented that allows different parts of the reservoir to be perturbed separately. This method improves the robustness, efficiency, and the convergence speed of the PPM. Three examples with synthetic, realistic and real training images are used to illustrate and validate the developed techniques for various types of reservoirs with different levels of geological complexities.Additionally, this work extends the applicability of MPS techniques for modelling some critical heterogeneities (i.e., complex fracture networks) present in unconventional tight and shale reservoirs. This extension allows us to apply the developed cluster-aided workflow for history matching the production data from multiple-fractured horizontal wells in future studies. Therefore, it is attempted to model complex fracture networks around primary hydraulic fractures using multiple-point geostatistical algorithms. Secondary probability maps, which can be derived from microseismic events, are also included in the modelling process to account for the variability in the extent of the induced and existing fracture networks within a stimulated reservoir volume near the primary hydraulic fractures. A sensitivity study is performed to understand the effect of different parameters on the well flow performance given different fracture network models. Optimization of subsurface flow processes can result in more economical projects. However, ignoring the uncertainty of geological realizations in the optimization process can lead to suboptimal outcomes that can be considerably different from the actual optimal solutions. A robust optimization workflow for the SAGD (steam assisted gravity drainage) process is presented that considers the geological uncertainties by optimizing over a subset of representative realizations obtained from three model ranking techniques. The first method is by applying the base case well locations and operating constraints to all the realizations and running simulations. The realizations are then ordered according to their resulting Net Present Values (NPVs). In the second ranking method, the clustering of a low-dimensional cumulative steam-oil ratio is employed as the feature vector for the SAGD process. Finally, the third method relies on kernel-clustering of permeability realizations using multidimensional scaling and similarity metrics. Although the best solution is obtained by the second ranking approach using the simulation results, the third method is more appealing as it provides competitive results using the static data only. | |
| dc.identifier.citation | Khani, H. (2020). Model Calibration and Performance Optimization Using Multiple-Point Geostatistics and Machine Learning Techniques (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. | |
| dc.identifier.uri | http://hdl.handle.net/1880/115480 | |
| dc.identifier.uri | https://dx.doi.org/10.11575/PRISM/40447 | |
| dc.language.iso | en | en |
| dc.language.iso | English | |
| dc.publisher.faculty | Graduate Studies | en |
| dc.publisher.faculty | Schulich School of Engineering | |
| dc.publisher.institution | University of Calgary | en |
| dc.rights | University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission. | en |
| dc.subject.classification | Engineering--Petroleum | |
| dc.subject.classification | Computer Science | |
| dc.subject.classification | Statistics | |
| dc.title | Model Calibration and Performance Optimization Using Multiple-Point Geostatistics and Machine Learning Techniques | |
| dc.type | doctoral thesis | |
| thesis.degree.discipline | Engineering – Chemical & Petroleum | |
| thesis.degree.grantor | University of Calgary | en |
| thesis.degree.grantor | University of Calgary | |
| thesis.degree.name | Doctor of Philosophy (PhD) |