Abstract:Abstract 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).

NMR log response prediction from conventional petrophysical logs with XGBoost-PSO framework

Permeability Prediction of Carbonate Reservoir Based on Nuclear Magnetic Resonance (NMR) Logging and Machine Learning

Approaches of Combining Machine Learning with NMR-Based Pore Structure Characterization for Reservoir Evaluation

A New Workflow for Estimating Reservoir Properties With Gradient Boosting Model and Joint Inversion Using MWD Measurements

Integration of electromagnetic, resistivity-based and production logging data for validating lithofacies and permeability predictive models with tree ensemble algorithms in heterogeneous carbonate reservoirs

Prediction of permeability of highly heterogeneous hydrocarbon reservoir from conventional petrophysical logs using optimized data-driven algorithms

Well-Logging Prediction Based on Hybrid Neural Network Model

NMR-Supported Near-Wellbore Data Analysis to Improve the Petrophysical Evaluation of a Sandstone Reservoir Drilled With Barite-Enriched Water-Based Mud

New Insight Into Data Processing of Nuclear Magnetic Resonance Logging for Pore Structure Characterization in Tight Sandstone Reservoirs

Permeability prediction of petroleum reservoirs using stochastic gradient boosting regression

Porosity prediction using ensemble machine learning approaches: A case study from Upper Assam basin

The Saturation Calculation of NMR Logging Based on Constructing Water Spectrum Function

A New Method for Mobility Logging Evaluation Based on Flowing Porosity in Shale Oil Reservoirs

Permeability Predictions for Tight Sandstone Reservoir Using Explainable Machine Learning and Particle Swarm Optimization

Shear wave velocity prediction based on adaptive particle swarm optimization optimized recurrent neural network

A nuclear magnetic resonance proxy model for predicting movable fluid of rocks based on adaptive ensemble learning

Deep learning for predicting permeability logs in offset wells using an artificial neural network at a Waggoner Ranch reservoir, Texas

Synthesize Nuclear Magnetic Resonance T2 Spectrum from Conventional Logging Responses with Spectrum Regression Forest.

Improving Permeability and Productivity Estimation with Electrofacies Classification and Core Data Collected in Multiple Oilfields

Logging-Based Petrophysical Estimation for Tight Sandy-Mud Reservoirs Employing a Geologically Regularized Learning System

Factors Affecting and Correction Methods for Porosity Measured by NMR Logging in the J Oilfield of Bohai Bay