Skip to main content

Über dieses Buch

This book presents several intelligent approaches for tackling and solving challenging practical problems facing those in the petroleum geosciences and petroleum industry. Written by experienced academics, this book offers state-of-the-art working examples and provides the reader with exposure to the latest developments in the field of intelligent methods applied to oil and gas research, exploration and production. It also analyzes the strengths and weaknesses of each method presented using benchmarking, whilst also emphasizing essential parameters such as robustness, accuracy, speed of convergence, computer time, overlearning and the role of normalization. The intelligent approaches presented include artificial neural networks, fuzzy logic, active learning method, genetic algorithms and support vector machines, amongst others.

Integration, handling data of immense size and uncertainty, and dealing with risk management are among crucial issues in petroleum geosciences. The problems we have to solve in this domain are becoming too complex to rely on a single discipline for effective solutions and the costs associated with poor predictions (e.g. dry holes) increase. Therefore, there is a need to establish a new approach aimed at proper integration of disciplines (such as petroleum engineering, geology, geophysics and geochemistry), data fusion, risk reduction and uncertainty management. These intelligent techniques can be used for uncertainty analysis, risk assessment, data fusion and mining, data analysis and interpretation, and knowledge discovery, from diverse data such as 3-D seismic, geological data, well logging, and production data. This book is intended for petroleum scientists, data miners, data scientists and professionals and post-graduate students involved in petroleum industry.



Intelligent Data Analysis Techniques—Machine Learning and Data Mining

This introductory chapter presents some of the main paradigms of intelligent data analysis provided by machine learning and data mining. After discussing several types of learning (supervised, unsupervised, semi-supervised, active and reinforcement learning) we examine several classes of learning algorithms (naive Bayes classifiers, decision trees, support vector machines, and neural networks) and the modalities to evaluate their performance. Examples of specific applications of algorithms are given using System R.
Dan Simovici

On Meta-heuristics in Optimization and Data Analysis. Application to Geosciences

This chapter presents popular meta-heuristics inspired from nature focusing on evolutionary computation (EC). The first section, as an elevator pitch, briefly walks through problem solving, touching upon notions such as optimization problems, meta-heuristics, constraint handling, hybridization, and the No Free Lunch Theorem for optimization, and also giving very short introductions into several most popular meta-heuristics. The next two sections are dedicated to evolutionary algorithms and swarm intelligence (SI), two of the main areas of EC. Three particular optimization methods illustrating these two areas are presented in more detail: genetic algorithms (GAs), differential evolution (DE), and particle swarm optimization (PSO). For a better understanding of these algorithms, references to R packages implementing the algorithms and code samples to solve numerical and combinatorial problems are given. The fourth section is dedicated to the use of EC techniques in data analysis. Optimization of the hyper-parameters of conventional machine learning techniques is illustrated by a case study. The last section reviews applications of meta-heuristics in geosciences.
Henri Luchian, Mihaela Elena Breaban, Andrei Bautu

Genetic Programming Techniques with Applications in the Oil and Gas Industry

The chapter, entitled “Genetic Programming Techniques with Applications in the Oil and Gas Industry”, consists of four parts. The first part presents theoretical features of the genetic programming algorithm, describing its main components, such as individual representation, initialization of the population, evaluation of the individuals, genetic operators, and selection scheme. The second part is concerned with a hybrid evolutionary algorithm—Gene Expression Programming, which combines features from genetic algorithms and genetic programming. In the third part, references towards software frameworks that implement GP are provided. This part then focuses on the use of the R package for genetic programming—RGP and provides a guide for the package, using two model problems to exemplify its usage. The last part reviews applications of genetic programming for petroleum engineering problems.
Henri Luchian, Andrei Băutu, Elena Băutu

Application of Artificial Neural Networks in Geoscience and Petroleum Industry

It has been shown that artificial neural networks (ANNs), as a method of artificial intelligence, have the potential to increase the ability of problem solving to geoscience and petroleum industry problems particularly in case of limited availability or lack of input data. ANN application has become widespread in engineering including geoscience and petroleum engineering because it has shown to be able to produce reasonable outputs for inputs it has not learned how to deal with. In this chapter, the following subjects are covered: artificial neural networks basics (neurons, activation function, ANN structure), feed-forward ANN, backpropagation and learning (perceptrons and backpropagation, multilayer ANNs and backpropagation algorithm), data processing by ANN (training, over-fitting, testing, validation), ANN and statistical parameters, an applied example of ANN, and applications of ANN in geoscience and petroleum Engineering.
Rahman Ashena, Gerhard Thonhauser

On Support Vector Regression to Predict Poisson’s Ratio and Young’s Modulus of Reservoir Rock

Accurate prediction of rock elastic properties is essential for wellbore stability analysis, hydraulic fracturing design, sand production prediction and management, and other geomechanical applications. The two most common required material properties are Poisson’s ratio and Young’s modulus. These elastic properties are often reliably determined from laboratory tests by using cores extracted from wells under simulated reservoir conditions. Unfortunately, most wells have limited core data. On the other hand, wells typically have log data. By using suitable regression models, the log data can be used to extend knowledge of core-based elastic properties to the entire field. Artificial neural networks (ANNs) have proven to be successful in many reservoir characterization problems. Although nonlinear problems can be well resolved by ANN-based models, extensive numerical experiments (training) must be done to optimize the network structure. In addition, generated regression models from ANNs may not perfectly generalize to unseen input data. Recently, support vector machines (SVMs) have proven successful in several real-world applications for its potential to generalize and converge to a global optimal solution. SVM models are based on the structural risk minimization principle that minimizes the generalization error by striking a balance between empirical training error and learning machine capacity. This has proven superior in several applications to the empirical risk minimization (ERM) principle adopted by ANNs that aims to reduce the training error only. Here, support vector regression (SVR) to predict Poisson’s ratio and Young’s modulus is described. The method uses a fuzzy-based ranking algorithm to select the most significant input variables and filter out dependency. The learning and predictive capabilities of the SVR method is compared to that of a backpropagation neural network (BPNN). The results demonstrate that SVR has similar or superior learning and prediction capabilities to that of the BPNN. Parameter sensitivity analysis was performed to investigate the effect of the SVM regularization parameter, the regression tube radius, and the type of kernel function used. The result shows that the capability of the SVM approximation depends strongly on these parameters.
A. F. Al-Anazi, I. D. Gates

Use of Active Learning Method to Determine the Presence and Estimate the Magnitude of Abnormally Pressured Fluid Zones: A Case Study from the Anadarko Basin, Oklahoma

We discuss active learning method (ALM) as an artificial intelligent approach for predicting a missing log (DT or sonic log) when only two other logs (GR and REID) are present. Applying ALM approach involves three steps: (1) supervised training of the model, using available GR, REID, and DT logs; (2) confirmation and validation of the model by blind-testing the results in a well containing both the predictors (GR, REID) and the target (DT) values; and (3) applying the predicted model to wells containing the predictor data and obtaining the synthetic (simulated) DT values. Our modeling approach indicates that the performance of the algorithm is satisfactory, while the time performance is significant. The quality of our simulation procedure was assessed by three parameters, namely mean square error (MSE), mean relative error (MRE), and Pearson product–momentum correlation coefficient (R). The values obtained for these three quality-control parameters appear congruent, with the exception of MRE, regardless of the training set used (reduced vs. complete). ALM performance was measured also by the time required to attain the desirable outcomes: five depth levels of investigation took a little more than one minute of computing time during which MSE dropped significantly. We performed twice the regression analysis: with and without normalization of input data sets (training well and validation well) using the procedure indicated by previous works. The results show minimum differences in quality assessment parameters (MSE, MRE, and R), suggesting that data normalization is not a necessary step in all regression algorithms. We employed both the measured and simulated sonic logs DT to predict the presence and estimate the depth intervals where overpressured fluid zone may develop in the Anadarko Basin, Oklahoma. Based on our interpretation of the sonic log trends, we inferred that overpressure regions are developing between ~1250 and 2500 m depth and the overpressured intervals have thicknesses varying between ~700 and 1000 m. These results match very well our previous results reported in the Anadarko Basin, using the same wells, but different artificial intelligent approaches.
Constantin Cranganu, Fouad Bahrpeyma

Active Learning Method for Estimating Missing Logs in Hydrocarbon Reservoirs

Characterization and estimation of physical properties are two of the most important key activities for successful exploration and exploitation in the petroleum industry. Pore-fluid pressures as well as estimating permeability, porosity, or fluid saturation are some of the important example of such activities. Due to various problems occurring during the measurements, e.g., incomplete logging, inappropriate data storage, or measurement errors, missing data maybe observed in recorded well logs. This unfortunate situation can be overcome by using soft computing approximation tools that will estimate the missing or incomplete data. Active learning method (ALM) is such a soft computing tool based on a recursive fuzzy modeling process meant to model multi-dimensional approximation problems. ALM breaks a multiple-input single-output system into some single-input single-output sub-systems and estimates the output by an interpolation. The heart of ALM is fuzzy measuring of the spread. In this paper, ALM is used to estimate missing logs in hydrocarbon reservoirs. The regression and normalized mean squared error (MSE) for estimating density log using ALM were equal to 0.9 and 0.042, respectively. The results, including errors and regression coefficients, proved that ALM was successful on processing the density estimation. ALM is illustrated by an example of a petroleum field in the NW Persian Gulf.
Fouad Bahrpeyma, Constantin Cranganu, Behrouz Zamani Dadaneh

Improving the Accuracy of Active Learning Method via Noise Injection for Estimating Hydraulic Flow Units: An Example from a Heterogeneous Carbonate Reservoir

Due to many reasons, in many occasions, reservoir engineers should analyze the reservoirs with small sets of measurements; this problem is known as the small sample size problem. Because of small sample size problem, modeling techniques commonly fail to accurately extract the true relationships between the inputs and the outputs used for reservoir properties prediction or modeling. In this paper, small sample size problem is addressed for modeling carbonate reservoirs by the active learning method (ALM). In this paper, noise injection technique, which is a popular solution to small sample size problem, is employed to recover the impact of separating the validation and test sets from the entire sample set in the process of ALM. The proposed method is used to model hydraulic flow units (HFUs). HFUs are defined as correlatable and mappable zones within a reservoir which control fluid flow. This study presents quantitative formulation between flow units and well logs data in one of the heterogeneous carbonate reservoir in Persian Gulf. The results for R and nMSE are equal to 85 % and 0.0042 which reflect the ability of the proposed method when facing with sample size problem.
Fouad Bahrpeyma, Constantin Cranganu, Bahman Golchin

Well Log Analysis by Global Optimization-based Interval Inversion Method

Artificial intelligence methods play an important role in solving an optimization problem in well log analysis. Global optimization procedures such as genetic algorithms and simulated annealing methods offer robust and highly accurate solution to several problems in petroleum geosciences. According to experience, these methods can be used effectively in the solution of well-logging inverse problems. Traditional inversion methods are used to process the borehole geophysical data collected at a given depth point. As having barely more types of probes than unknowns in a given depth, a set of marginally over-determined inverse problems has to be solved along a borehole. This single inversion scheme represents a relatively noise-sensitive interpretation procedure. For the reduction of noise, the degree of over-determination of the inverse problem must be increased. To fulfill this requirement, the so-called interval inversion method is developed, which inverts all data from a greater depth interval jointly to estimate petrophysical parameters of hydrocarbon reservoirs to the same interval. The chapter gives a detailed description of the interval inversion problem, which is solved by a series expansion-based discretization technique. Different types of basis functions can be used in series expansion depending on the geological structure to treat much more data against unknowns. The high degree of over-determination significantly increases the accuracy of parameter estimation. The quality improvement in the accuracy of estimated model parameters often leads to a more reliable calculation of hydrocarbon reserves. The knowledge of formation boundaries is also required for reserve calculation. Well logs do contain information about layer thicknesses, which cannot be extracted by the traditional local inversion approach. The interval inversion method is applicable to derive the layer boundary coordinates and certain zone parameters involved in the interpretation problem automatically. In this chapter, it is analyzed how to apply a fully automated procedure for the determination of rock interfaces and petrophysical parameters of hydrocarbon formations. Cluster analysis of well-logging data is performed as a preliminary data processing step before inversion. The analysis of cluster number log allows the separation of formations and gives an initial estimate for layer thicknesses. In the global inversion phase, the model including petrophysical parameters and layer boundary coordinates is progressively refined to achieve an optimal solution. The very fast simulated re-annealing method ensures the best fit between the measured data and theoretical data calculated on the model. The inversion methodology is demonstrated by a hydrocarbon field example, which shows an application for shaly sand reservoirs. The theoretical part of the chapter gives a detailed mathematical formulation of the inverse problem, while the case study focuses on the practical details of its solution by using artificial intelligence tools.
Mihály Dobróka, Norbert Péter Szabó

Permeability Estimation in Petroleum Reservoir by Meta-heuristics: An Overview

Proper permeability distribution in reservoir models is very important in the determination of oil and gas reservoir quality. In fact, it is not possible to have accurate solutions in many petroleum engineering problems without having accurate values for this key parameter of hydrocarbon reservoir. Permeability estimation by individual techniques within the various porous media can vary with the state of in situ environment, fluid distribution, and the scale of the medium under investigation. Recently, attempts have been made to utilize artificial intelligent methods for the identification of the relationship which may exist between the well log data and core permeability. This study overviews the different artificial intelligent methods in permeability prediction with advantage of each method. Finally, some suggestions and comments to choose the best method are introduced.
Ali Mohebbi, Hossein Kaydani


Weitere Informationen