Célio Maschio

A new upscaling technique based on Dykstra-Parsons coefficient : Evaluation with streamline reservoir simulation

Abstract The loss of information is inevitable in any upscaling technique. The efficiency of a given method must take in account two basic aspects: the first is the agreement of the results obtained with the coarse grid when compared to the results obtained with the fine grid and the second is the upscaling computational performance. Upscaling is really justifiable for very fine-scale reservoir models, usually with more than 1 million blocks or when many runs are necessary and the computational performance is very important. Due to the size of such fine-scale models, two problems exist. The first problem is that application of a pressure solver technique (or numerical method, which is generally more appropriate) on the upscaling is very time-consuming. The second problem is that the flux simulation of the fine-scale model, in order to validate a given upscaling technique, is very difficult and sometimes impossible through traditional simulators. In this work, an upscaling technique based on a heterogeneity coefficient (Dykstra–Parsons), which is as efficient and faster than numerical methods, is proposed. All tested fine grid reservoir models were modeled by a streamline simulator. The proposed technique was applied to three case studies obtaining good agreement between coarse and fine grids, taking in account several production parameters.

Application of Neural Network and Global Optimization in History Matching

This article presents the application of global optimizers combined with Artificial Neural Networks (ANN) to the history matching problem. An evolutionary algorithm is executed on the proxy generated through the ANN technique. The results obtained from evolutionary algorithms are fine-tuned by using a local optimizer based on the Hooke and Jeeves optimization method. The methodology is applied in two reservoir models and promising results were obtained.

Modeling of the efficiency of fibrous filters through numerical simulation and X-ray tomography

This paper describes modeling of the efficiency of fibrous filters using numerical simulation and X-ray computerized tomography (XCT). Variations in the flow field within the filter due to particle accumulation and structural inhomogeneities are taken into account in the mathematical model. Both the theoretical results and the experimental XCT data (digital images) showed that in the depth filtration process analyzed there is graded particle accumulation. Larger particles form bridges and can improve the collection for smaller particles during the filtration time. It is also found that smaller particles tend to promote a more pronounced pressure drop.

Scatter Search Metaheuristic Applied to the History-Matching Problem

Reservoir simulation plays an important role in man agerial decisions during the life of a reservoir, so it is v tal that they present an accurate prospect of real reservoir perf ormance. The main goal of the History Matching (HM) process i to improve the quality of numerical models by constrai ning simulated to observed data. A typical HM process ev aluates a chosen objective function (OF) comprised by linear functions on the uncertain reservoir properties. The OF can b e modeled in such a way that the HM process can be solved as an optimization problem. This paper discusses how the Scatter Search (SS) technique can solve the HM problem. The main feature of SS is that it works on a set of solution s called the reference set (RefSet). The idea is to improve the ov rall quality of the RefSet. New solutions are generated by a nonconvex combination of explored solutions. The goal f this paper is solve the HM with SS and to evaluate its performance. The proposed methodology was tested with two base synthetic reservoir models. The first is a homogeno us reservoir with 8 different horizontal permeability regions, while the second is a highly heterogeneous reservoi r m del where low quality background sand is crossed by hig h permeability canals. The results show that SS was q uite efficient, considering the quality of the generated solutions and the number of required numerical simulations. Most of the current HM methodologies do not perform well when the solution space is large and complex. The application of the SS methodology to the HM problem is a novel approach. Unlike most metaheuristics, SS can be effective even when the simulation time of each ten tative solution to the problem is long. Introduction Reservoir simulation has established itself as a ce ntral tool in studying different scenarios by which a petroleum r ese voir might be developed. Since so many decisions are tak en based on numerical simulation results, it is very importa nt that the simulation gives a good approximation of the perfor mance of the real reservoir. History Matching (HM) is the pr ocess of making changes on the properties of the reservoir m del so its simulation results better mirror real production/in jection data, in the expectation that the resulting model will al so yield good performance forecasts. The fact is that HM is, by nature, an ill posed inv erse problem which cannot be solved analytically, has a high degree of non-linearity, is usually comprised of a large number of variables, and, in most cases, accepts multiple so utions. These and other reasons make the HM problem one of the most time consuming phases on a reservoir simulatio n study, accounting for an average of one-third of the total study time [1]. Manual HM brings a lot of practical challenges with . There is always a large amount of data involved, ma ny simulation runs are needed, several comparisons bet ween simulated and observed data are required and manual modifications on the reservoir model are performed. All in all, it’s an error prone process. It is not likely that an efficient, fully automatic HM process is developed anytime soon because there are a lot of steps on the process which are subjective. The main subjective step is parameterization, i.e., which reservoir par ameters should be manipulated to tilt the simulated perform ance in the “right” direction. This part of the HM process rema ins in the realm of the reservoir engineer. Other tasks, howev er, are more repetitive, tedious and thus, subject to autom a ion. Once a set of reservoir parameters are chosen along with their variation intervals, the generation of reservoir mo dels containing different combinations of parameter valu es and the evaluation of how close a match results are tasks w hich can and should be automated to decrease the time spent in HM. This paper outlines how the HM problem can be model ed as a Combinatorial Optimization (CO) problem so it can be solved by meta-heuristic methods like Scatter Searc h (SS). After the SS implementation is described, two reser voir simulation models are adjusted using the proposed methodology. Optimization Background This section introduces a few terms and concepts co mm n in the optimization field which are important to bette r understand this work. SPE 102975 Scatter Search Metaheuristic Applied to the History-Matching Problem Sergio. H. G. Sousa, SPE, Célio Maschio, SPE, and Denis. J. Schiozer, SPE, UNICAMP

Quantifying the Impact of Grid Size, Upscaling, and Streamline Simulation in the Risk Analysis Applied to Petroleum Field Development

Geological uncertainties usually have a strong impact on the decisions applied to petroleum field development. Methodologies to quantify the impact of uncertainties are still not well established due to the amount of variables that have to be considered. The complete analysis usually depends on geological, economic and technological uncertainties that have different degrees of impact in the recovery process and may affect the decision process in different levels depending on the problem, reservoir characteristics, recovery mechanism and stage of field development. The objective of this paper is to evaluate the importance of the numerical simulation process in a risk analysis applied to a petroleum field development. The use of fast estimation of recovery factor can be used in the exploration risk analysis but, during appraisal and development phases, more reliable techniques must be used to predict the reservoir performance in order to quantify accurately the impact of uncertainties and evaluate risk. One of the drawbacks of the use of numerical simulations can be the required computational effort. Therefore, this paper evaluates the impact of using different degrees of accuracy in the procedure, calculating risk with different options of grid size, number of attributes, levels of uncertainties and also comparing the results with streamline simulation which is used also for very fine models. The proposed procedure is tested in the SPE 10th Comparative Solution and a comparison is presented for several different options. Finally some conclusions are taken based on accuracy, required computational effort, model scale, comparing also results from simulation through finite difference model and streamline. Introduction In the past, prediction of reservoir performance and field economic evaluation were normally carried out with one or a few possible models mainly due to restrictions in the computer resources. Nowadays, probabilistic forecasts are possible but due to the high number of uncertain variables and complexity of the models, simplifications are still necessary. Petroleum field development is always strongly related to uncertainties and risk. The complete economic analysis of the problem involves many uncertain parameters with different degrees of impact on the decisions that have to taken in the process. The most important uncertainties are related to geological models, economic parameters and technology. To evaluate the risk involved in the process, it is necessary to quantify the impact of these uncertainties, which is normally measured with objective functions. The analysis can be simplified if the uncertain parameters are combined in three groups: volumes in place (VHIP), recovery factor (RF) and economic model. Uncertainty related to volumes in place is a direct function of the geological characterization and it is normally reduced significantly during appraisal phase although it also may be responsible by high risk variations due to new and sometimes unexpected information. Considering the economic model, oil price is the major font of uncertainty. Investments and costs variations related to new technologies have also to be considered due to the long period of typical projects, especially in offshore fields. Regulatory issues also may affect in some cases. All parameters that affect reservoir behavior may be considered in the uncertainties related to recovery factors. In such cases, rock and fluid attributes, reservoir mechanism and production strategy are the most common parameters to be considered. These uncertainties are more important during appraisal and development phases when many important decisions related to production facilities are taken. One of the main goals of the risk evaluation is to identify the impact of each uncertainty in order to simplify the problem without significant accuracy reduction. The problem becomes more difficult because the impact of these uncertainties varies during the development of the fields. Most of the published works (Newendorp, 1975; Garb, 1988) related to risk measurement have the focus on the exploration phase where uncertainties due to reservoir performance prediction have small impact and probabilistic SPE 79677 Quantifying the Impact of Grid Size, Upscaling, and Streamline Simulation in the Risk Analysis Applied to Petroleum Field Development Eliana L. Ligero, SPE, UNICAMP; Célio Maschio, SPE, UNICAMP; Denis J. Schiozer, SPE, UNICAMP

Risk Assessment of Petroleum Fields—Use of Numerical Simulation and Proxy Models

Abstract The development of petroleum fields is a complex task due to the high influence of uncertainties on E&P projects. During the appraisal and development phases, uncertainties related to geologic and fluid models play an important role, especially in offshore heavy oil fields due to the low economic return, limited flexibility, and importance of reservoir modeling. The flexibility is limited because of the necessity to design the production facilities based on a low amount of information. The reservoir modeling process is important because risk of field development projects is normally caused by a high uncertainty on the recovery factor. Due to the necessity of a more robust evaluation of recovery factor, risk assessment methodologies normally are integrated with reservoir simulation, which is the best available tool to predict reservoir performance. However, higher precision on prediction of reservoir behavior is normally associated with fine simulation grid and high computation effort. In this article, some alternatives are presented to improve the efficiency of risk assessment, considering precision and computation effort. Among these alternatives are (1) use of coarse models, (2) use of coarse models modified to reproduce fine grid results, (3) simplifications on the risk assessment procedure, and (4) use of proxy models based on statistical (experimental) design and response surface methodology. A general discussion, including each alternative, use of upscaling techniques, reduction of grid size, number of attributes, use of parallel computing, and use of proxy models are made based on previous publications and results of a case study. The methodology applied to quantify risk involves a sensitivity analysis in order to reduce the number of critical attributes and simulation of reservoir models obtained through the combination of these attributes. Afterward, a statistic treatment is used to evaluate the risk involved in the process. Based on a case study, it is shown that the use of faster simulation models and proxies can speed up risk assessment, but a few steps must be performed to guarantee the quality of the results.

A New Methodology for Assisted History Matching Using Independent Objective Functions

Abstract History matching is a complex inverse problem for which the degree of difficulty and the computational effort (in terms of number of simulations) increase with the increasing of the number of matching parameters. This article presents a new methodology for assisted history matching based on independent objective functions that decrease the number of simulations. The proposed approach consists of the optimization of several objective functions related to each region of the reservoir to be matched, such as a well or a group of wells. Optimization processes, one for each objective function, are started simultaneously, modifying the same data file, yielding a more efficient process, allowing speedup and preserving the quality of the results. The methodology was successfully applied to an offshore field. The results show that the quality of results is practically the same when compared to the conventional procedures, i.e., matching of the wells individually or combining several wells. The advantage is a significant reduction on the number of simulations, preserving the quality of the results.

Pressure and saturation estimation from P and S impedances: a theoretical study

This work represents the first step of a study to integrate time lapse seismic and reservoir engineering data where a petro-elastic inversion from seismic data to pressure and saturation is presented. This inversion is made through an optimization procedure. In order to better understand and validate the initial step of the methodology, synthetic data (initially free of noise and errors) have been used. Through this ideal set of data, it was possible to show that pressure and saturation can be extracted from P and S impedances using only one seismic survey (3D inversion). It is also shown that this 3D approach is not robust when errors are assumed in reservoir data and it fails when, for instance, uncertainty in porosity data occurs. Thus, an improvement is made and the algorithm is rewritten based on 4D differences that diminish the wrong reservoir data effect. For both algorithms (3D and 4D), we have presented a discussion of the objective function behaviour concerning the use of P and S impedances simultaneously, the initial guess and the solution space. A sensitivity analysis discussing the influence of porosity and the dynamic properties on P and S impedances for 3D and 4D approaches is also presented. After understanding the inversion process behaviour for an ideal data set, an analysis of its results assuming different combinations of pressure and saturation variations and including some errors in the data set used is presented in the last subsections.

Integration of History Matching And Uncertainty Analysis

The most common procedure to perform a production history matching is to start with a base model and modify reservoir and fluid properties to adjust simulation results with the production history of the field. This paper presents an example of a different procedure. The history matching starts with an uncertainty analysis where several possible models are generated and the models that do not reproduce the behaviour of the reservoir are discarded: 1) allowing a faster history matching process; 2) increasing confidence in the process; and, 3) adding uncertainty analysis to the production prediction. The methodology starts with a dynamic sensitivity analysis based on simulation of models where uncertain attributes are tested and compared with a base model. The attributes are then selected and combined in a derivative tree. Several models are evaluated and those that do not match the production and pressure history are discarded, reducing uncertainty in prediction of the behaviour of the reservoir. This methodology was motivated by a reservoir with unexpected behaviour that yielded a difficult history matching when considering the usual procedures. Techniques were developed to analyze reservoir performance and differential pressure maps between zones. New approaches to assess connectivity between zones were used to give alternate structural models to the sensitivity analysis. This methodology may be helpful during the first years of production when uncertainties are still significant and when typical procedures to perform production history matching are unable to solve the problem efficiently.

A new optimization framework using genetic algorithm and artificial neural network to reduce uncertainties in petroleum reservoir models

In this article, a new optimization framework to reduce uncertainties in petroleum reservoir attributes using artificial intelligence techniques (neural network and genetic algorithm) is proposed. Instead of using the deterministic values of the reservoir properties, as in a conventional process, the parameters of the probability density function of each uncertain attribute are set as design variables in an optimization process using a genetic algorithm. The objective function (OF) is based on the misfit of a set of models, sampled from the probability density function, and a symmetry factor (which represents the distribution of curves around the history) is used as weight in the OF. Artificial neural networks are trained to represent the production curves of each well and the proxy models generated are used to evaluate the OF in the optimization process. The proposed method was applied to a reservoir with 16 uncertain attributes and promising results were obtained.