Numerical tuning in reservoir simulation: it is worth the effort in practical petroleum applications

Abstract

Giant carbonate oil fields in Brazilian pre-salt area present a high level of heterogeneity. Despite the fact that geoscientists aim at modelling these complex models in a high-block cell resolution, it is important to reduce the computational effort and speed up some processes such as forecasting the production during a risk analysis in a probabilistic approach. This practice must be evaluated to keep numerical and geological consistency of reservoir models. As a result, creating a procedure to assist petroleum scientists and engineers so as to reduce computational time, without losing accuracy, is the main goal of this proposed work. A methodology based on numerical parameter evaluation, tuning technique, diagnosis and application is defined to provide a robust and effective starting point of the numerical control settings during reservoir simulation model runs. All of the analyses are based on a good understanding of the reservoir heterogeneities (geologists) and characteristics (reservoir engineering) that may result in reduced central processing unit time, number of time-step cuts, solver failures and material balance compared to the default approach considered for these problems. In addition, to check the consistency of the proposed procedure, one building risk curve application for two case studies was selected, assuming that we are executing the tuning and the application in sequence of a common reservoir application. The results of the two case applications under our proposed assumptions showed that it is possible to (1) speed the simulation runs up to three times in comparison with the base case which uses the default numerical parameters of a commercial simulator, (2) reduce the number of cuts and failures and (3) control the material balance equation error within a previously defined tolerance. Despite decreasing the simulation run time in practical reservoir study applications, we are also avoiding bad combination of numerical inputs which can result in unfeasible reservoir numerical models. To conclude, depending on the amount of applications, number of runs and complexity of reservoir numerical model, the results showed that we can save time in the reservoir engineer's routine activities while maintaining the consistency of reservoir models by dedicating an initial time to better understand and optimise the numerical parameters. Even though the vast majority of reservoir numerical models are created in highresolution grid cells to maintain the level of heterogeneity and, consequently, increase the computational effort, more suitable numerical model quantifications should be conducted in order to reduce the running time. Nonetheless, a consistency in the reservoir model must be maintained during the entire procedure. Thus, the most important step before starting any reservoir simulation workflow is to set the numerical parameters to avoid convergence problems, to keep the consistency of the numerical simulation results (approaching the solution of nonlinear equations to the true solution) and to ass...

Model-based decision analysis applied to petroleum field development and management

“…2. Build and calibrate the simulation model: accurate risk quantification requires reliable responses; therefore, the simulation model must be calibrated to have a fast and yet robust response to avoid biased evaluations (Avansi et al, 2019). Decision makers define the degree of model precision according to the objective.…”

TunaOil: A Tuning Algorithm Strategy for Reservoir Simulation Workloads

“…In [16], the authors apply the DECE (Designed Exploration and Controlled Evolution) [17] method twice successively. The first performs a random level selection for each parameter with a Tabu Search [18] and Experimental Design [19], while the second optimizes the objective function (the elapsed time in the case).…”

“…They compare the performance of the numerical tuning against the default parameters of two related models: UNISIM-I-D-h7v7 (an 81x58x23 corner-point grid model) and UNISIM-I-D-hfm (a 326x234x157 corner-point grid model, almost 10x the size of the smaller one in terms of total cells). Their search space consisted of the same 12 parameters used in work [16]. They did not consider parallelization as one of the parameters and used only one or two processors in the evaluation.…”

A Review of Proxy Modeling Highlighting Applications for Reservoir Engineering

“…Performing an analysis with a low number of cells might be quick; however, it sacrifices the quality of the results, or it does not yield convergence. Conversely, a high number of cells increases the computational time, so obtaining the results at the various realizations of the problem is very time-consuming [5]. In recent years, improvements in computational hardware and software, and the emergence of the parallel processing of CPUs have boosted the speed of running numerical models.…”