Influence of the objective function in the history matching process

An assisted history matching process always requires two distinct parts: a parameterization to select the uncertain attributes of the model and an automatic procedure that minimizes the distance between the observed production and the simulation model curves. The focus of this work is the second part, where an objective function is necessary to represent mathematically the quality of the model. However, due to the complexity of the models, this function is frequently a combination of several functions that represent the quality of the match in several wells and less attention is given to the influence of the objective function in the optimization process. This paper proposes a study to show the influence of a global objective function on the history matching process using a synthetic reservoir model with 20 uncertain attributes. Results of the quality matching index of eight different global objective functions are compared at the end of the process. The optimized simulation models, generated by the optimization phase with different global objective functions, are compared with the base model and production history.

Research highlights
► Methodology uses different objective functions to test the matching performance.
► The sequential quadratic programming was the chosen optimization process.
► We examine a case with 20 uncertain attributes with multiple combinations.
► Results show the influence of the objective function in the history matching.