Regis Kruel Romeu

Novel Ways of Parameterizing the History-Matching Problem

One of the challenges when making a history match study is to find an adequate parameterization for the reservoir model. The main assumptions of the geological characterization should be respect and the influence of parameters on the fluid flow simulation results should be taken into account. On the other hand, the number of parameters should be kept within reasonable bounds in order to make the process viable. In this work, three examples of novel ways to parameterize the history match problem will be shown. Two of them are real field cases and one is a synthetic case based on outcrop data. Common to all examples is the choice of parameters that are related to the geological model building process, such as the variogram in a geostatiscal modeling or correlations between petrophysical properties (permeability x porosity, for instance). In this context, the use of a versatile history matching tool was essential, allowing for a quantitative evaluation for the quality of the match and for managing a larger number of parameters, when comparing to the traditional trial and error procedure. These examples show how the combination of a suitable parameterization with a versatile assisted history matching tool can improve both the quality and the efficiency of the history matching process. Introduction History matching is the process of modifying parameters in a reservoir simulation model, such as permeability and porosity, in order to make the model fit the production data previously observed in the field. Traditionally, this is pursued through a long and tedious process of trial and error, involving a qualitative judgment of the match. In general, the modifications in the model do not fully respect the assumptions made in the geological model building process, so that static and dynamic data are not incorporated in a systematic way. Also, many different solutions are possible, depending on the starting model and the reservoir parameters that will be modified. Due to these reasons, the history matching is usually the most complex and time-consuming task in a reservoir simulation study, requiring a lot of expertise and experience. These difficulties have long been recognized and many techniques aiming to automate the process have been presented in the literature and, in the last few years, some assisted history matching tools have become commercially available. Common to most of these methodologies is the use of a least-squares objective function to quantify the misfit between simulated and observed data. This objective function can be expressed as

A multi-scale approach to improve reservoir characterization and forecasting: the Albacora Field (deep-water offshore Brazil) study

The Upper Albian Namorado Sandstone is one of the reservoirs of the Albacora Field, located in the Campos Basin, deep-water offshore Brazil. It is a sand-rich turbidite system where the most important controls on permeability are calcite cementation, thin beds of non-reservoir lithologies and some north–south trending faults. A major multidisciplinary reservoir characterization project was conducted to improve the reservoir description using all available data. In this paper, we focus on how the effect of rock heterogeneities were represented in the fluid flow model and on the performance obtained from this model. The basic idea was to define a hierarchical model of facies established on the basis of three main work scales: porous systems (thin sections and core sample scale); composite facies (whole core and log scale); and seismic facies (interwell to field scale). An up-scaling technique, based on the geopseudo concept, was used to generate the effective petrophysical properties for the fluid-flow simulation model. A Markov–Bayes geostatistical simulation method was applied in facies stochastic modelling. The sophisticated model that was built allowed very fast history matching.

Reservoir Representation for Flow Simulation

Reservoir flow modeling involves two aspects: a functional model (flow equations and numerical methods) and a representation model (related to the reservoir description) – and this last aspect is usually the critical one. Reservoir representation is different from reservoir characterization. It is not a question of describing the reservoir in a more or less exhaustive or realistic way, but a question of incorporating relevant information into the flow simulation model, considering the syntax of the flow simulators and the relative impact of information on the simulation results. Reservoir representation makes the bridge between reservoir characterization and flow simulation. This paper presents selected results from a cluster of applied research projects in reservoir representation for flow simulation. It includes: (1) gridding issues in the context of integrated reservoir studies (discussion of terminology, grid specification, generic format reading, etc.); (2) identification and representation of critical heterogeneities of turbidite reservoirs; (3) assigning transmissibility multipliers across partially sealing faults in a field case; (4) incorporation of production data in reservoir flow models with an example of application; (5) hyperdocumention for reservoir simulation files (making use of HTML tags to get a much richer documentation of the flow simulation model).

Flow Simulation Study of the Namorado Sandstone (Albacora Field, offshore Brazil) Accounting for Scaling of Petrophysical Properties

This paper describes the reservoir flow simulation study of the Namorado Sandstone, Albacora Field (offshore Brazil). This study is the end point activity - to which the geological description, geostatistical modeling and upscaling results converge - of the project PRAVAP-2, a main project designed to develop technology on reservoir characterization at Petrobras. This paper details work executed by a multidisciplinary team on a flow simulation study performed in a stochastic modeling scenario. We present aspects related to the model construction and the results obtained from the flow simulation study. The fluid flow model considers a deterministic framework built with high-resolution stratigraphy and detailed structural analysis. Twelve stratigraphic units were grouped in six simulation layers. The heterogeneities inside each layer were modeled using a geostatistical approach: we used the MarkovBayes simulation method to integrate seismic and well data. A multi-step upscaling technique was used to generate the effective petrophysical properties for the fluid flow simulation model. The idea is based on (1) a hierarchical description of the reservoir heterogeneities and on (2) the computation of the effective properties bottom-up through this hierarchy. Three main work scales were defined: “porous systems” (microscopic scale); “composite facies” (log and core scale); and “seismic facies” (interwell to field scale). Effective permeabilities and two-phase pseudo-functions were numerically computed for each facies at each scale. The simulation grid has a corner point geometry, comprising a total of 46,200 gridblocks with areal dimensions of 100m x 100m. The interconnection of layers was modeled by changing the vertical transmissibility distribution. The history matching was performed by the adjustment of the average pressure for the six layers and of the reservoir production history. The results demonstrated that our sophisticated model allowed a very fast history matching. Extrapolations investigated the adequacy of the development plan, in special the type of the well, horizontal or hydraulic fractured vertical, the number and locations of these development wells to be drilled.