Field Study Application of Ensemble Based Assisted History Matching and Optimization for Reservoir Management

Abstract

Abstract Design of an optimal reservoir management strategy requires reliable reservoir performance prediction by the simulation models as well as implementation of robust and efficient mathematical optimization techniques to optimize the operating production and injection well controls. This paper describes an overall reservoir management workflow which consists of (1) the application of a novel data assimilation method for assisted history matching, and (2) the optimization of water-alternating-gas (WAG) injection cycles to maximize the life-cycle estimated ultimate oil recovery (EUR) of the reservoir. In this field study, the permeability field of the reservoir model is tuned to match the historical data. An iterative ensemble smoother (iES) optimization algorithm is used for the tuning, with a parameterization method based on coarsening of the property fields. The observed data includes the production and injection streams, the well bottom-hole pressures and static pressure measurements. The proposed parameterization combined with the iES algorithm resulted in a near-perfect match of the historical data. The property field tuning was within the acceptable and prior uncertainty ranges. For the WAG injection optimization study, the field management strategy and the list of WAG wells were given and fixed. The strategy included target average reservoir pressure, voidage replacement ratio and fluid material balance, in addition to the individual well limits and targets. The study parameters were water and gas injection cycle length. A design of experiment was done to understand the range of EUR for the field and understand its sensitivity to the optimization parameters. Next, a stochastic optimization algorithm was utilized to optimize the life-cycle EUR of the field by adjusting the injection cycle's length. The optimization was successful in improving the EUR of the field - the optimized ultimate recovery outperformed the best engineering design case and the ensemble of Latin Hypercube design cases. That said, the underlying field management strategy seemed to moderate the input controls, hence, the ranges of EUR improvements were relatively small. This study presents a successful application of novel ensemble based algorithms for field management and optimization, where the workflows outperformed the manual solutions both in terms of quality as well as the amount of effort required to do the study. The presented ensemble methods were efficient for estimating the sensitivities required for optimization of large scale problems - a critical aspect for field applications with large and computationally expensive simulation models.