Video: Improving Permeability and Productivity Estimation with Electrofacies Classification and Core Data Collected in Multiple Oilfields
- Xinlei Shi (CNOOC China Ltd. Tianjin) | Hongbing Chen (CNOOC China Ltd. Tianjin) | Ruijuan Li (CNOOC China Ltd. Tianjin) | Xiaoyan Yang (CNOOC China Ltd. Tianjin) | Huan Liu (CNOOC China Ltd. Tianjin) | Ting Li (Schlumberger)
- Document ID
- Offshore Technology Conference
- Publication Date
- Document Type
- 2019. Copyright is retained by the author. This presentation is distributed by OTC with the permission of the author. Contact the author for permission to use material from this video.
- 7.6.6 Artificial Intelligence, 5.6.2 Core Analysis, 5.6.4 Drillstem/Well Testing, 5.6.1 Open hole/cased hole log analysis, 3 Production and Well Operations, 5 Reservoir Desciption & Dynamics, 5.5.2 Core Analysis, 1.6.9 Coring, Fishing, 7 Management and Information, 5.6 Formation Evaluation & Management, 7.6 Information Management and Systems, 1.10 Drilling Equipment, 1.10 Drilling Equipment, 7.6.7 Neural Networks, 1.6 Drilling Operations
- Permeability, Core data, Multiple Oilfields, Productivity index, ElectroFacies Classification
- 0 in the last 30 days
- 1 since 2007
- Show more detail
- View rights & permissions
In the industry, it is a common practice to estimate continuous permeability by establishing a porosity-permeability relationship (poroperm) from conventional core analysis. For each new oilfield, core data is required to build a permeability model for this particular field. Due to reservoir heterogeneity, core derived poroperm can sometimes lead to biased predictions. This is particularly true for oilfields where core samples are scarce or provide a poor coverage of the reservoirs. Improving the accuracy of permeability models in these oilfields is key to better productivity estimation in the oilfield development planning.
In 1984, Hearn et al. first proposed the concept of flow unit while studying Shannon reservoir in HartogDraw oilfield, Wyoming, USA. Since Hearn put forward the concept of reservoir flow unit, various Electrofacies classification methods have been proposed by different scholars (Hearn et al. 1984). Generally they can be divided into two categories. One is geological research method, which mainly uses geological cuttings and routine core analysis to calculate flow zone index (FZI) for reservoir classification (Xinlei et al. 2017; Elphick et al. 1999; Kohonen et al. 1982). This method improves the accuracy of permeability evaluation to a certain extent, but it mainly relies on routine core analysis data. Due to poor ductility, this method has certain limitations in the classification of uncored reservoirs. The other is the relatively popular artificial intelligence technology in the oil industry in recent years. With the rapid development of computer hardware, artificial intelligence as a new technology is becoming more and more popular. In particular, the machine learning algorithm represented by neural network has a long history in petroleum industry technology, which solves many problems in petroleum specialty and is favored by many petroleum engineers. Machine learning classifies electrofacies mainly by clustering analysis of logging curves through mathematical algorithms such as neural network classification, K-nearest neighbor classification (KNN) and Multi-Resolution Graph based Clustering (MRGC), and then the corresponding relationship between electrofacies and lithofacies is established by combining core analysis and cutting data. Since this method is based on continuous well logs, it has strong extensibility and is easy to learn from uncored wells (Xinlei et al. 2017).
In this paper, we describe a novel workflow that predicts continuous permeability from conventional well logs, based on Electrofacies classification and core data collected in multiple oilfields. In this method, firstly, the MRGC is used to classify electrofacies of the logging curves in coring sections. Secondly, KNN algorithm is used to learn the results of electrofacies classification into uncored sections. Finally, the permeability model based on the electrofacies constraint is established. Compared with the neural network classification, the MRGC has the advantages of fast operation speed and stable operation results. The Neighbor Index (NI) parameter in the algorithm can quickly classify the sample data, and the Kernel Representative Index (KRI) parameter can select the optimal class from the results of multiple classifications(Yunjiang et al. 2018; Ting et al. 2018). Our study area consists of 13 oilfields with the same depositional environment and mineralogy. As a result, well log responses in these oilfields have similar characteristics. A total of 2122 core samples were collected in these oilfields and triple combo well logs are also available in the same wells.
Based on routine core analysis and log feature analysis, we divide log responses into 6 electrofacies. Permeability models are then established for each electrofacies using core data and are used to make predictions in new wells without any core data. Using the proposed idea, we re-estimated the permeability and productivity in a producer well in the study area. The facies constrained permeability shows a much better match with core measurement compared to conventional methods. As a result of the improved permeability, the productivity index calculated by the workflow matches that estimated by the Drill Stem Test (DST).