From Stochastic Geological Description to Production Forecasting in Heterogeneous Layered Reservoirs

Abstract

Abstract This paper discusses the data transfer from a detailed gridded stochastic geological model to a more coarsely gridded reservoir simulation model. A new gridding technique has been developed to follow the main geological features. This technique makes the transformation between the geological model and the reservoir model more accurate and efficient. It has been successfully applied in field evaluations for a North Sea sandstone reservoir. The reservoir in question is very heterogeneous. Conventional geological models based on smooth parameter interpolation between well data do not reflect these heterogeneities. A two-stage stochastic geological model was therefore used to represent the reservoir. In the first stage the geological architecture of the reservoir was constructed. In the second stage petrophysical values were assigned to each building block in the geological model. This two-stage modelling procedure was described by Damsleth et. al. The complete study is reported, from the stochastic geological model via the new gridding procedure and homogenization to simulated production forecasts. The effects of different assumptions in the gridding have been evaluated using both water, gas and WAG injection as well as depletion. A procedure for incorporating the results into the total field production forecast is outlined. The importance of representing the main geological features in the gridding is demonstrated. When a regular coarse grid was used, the contrasts in reservoir properties were smoothed out by averaging, and in most cases a more optimistic production performance was then predicted. Detailed stochastic geological models can give more realistic production profiles for the reservoir. In this study the stochastic model resulted in a reduction in recovery compared to conventional models. Introduction In the North Sea, only a few exploration and appraisal wells can be economically justified before important field development decisions have to be made. Due to incomplete reservoir information, oversimplified geological models are often constructed and used in simulation models. The use of such models based on data from a limited number of widely spaced wells has lead to many failures in predicting field performances. During the last ten years a variety of different methods for detailed stochastic geological modelling has been presented. Haldorsen and Damsleth list the following reasons for performing stochastic modelling:incomplete information about the reservoir;complex spatial disposition of reservoir building blocks or facies;difficult to capture rock-properties with spatial position and direction;the unknown relationships between the rock property value and the volume of rock used for averaging;the relative abundance of static over dynamic reservoir data; andconvenience and speed. However, the data transfer from the resulting geological models to more coarsely gridded reservoir simulation models has normally been inefficient and inaccurate. P. 311^