D-CRMP history matching considering predictive properties

Abstract

The article presents results of tests of software that implements the D-CRMP model. D-CRMP is a version of the analytical capacitance-resistance model (CRM) that is primarily used for waterflood characterization and reservoir management. The difference of D-CRMP lies in its ability to take into account the shut-in periods of production wells during history matching. The optimization problem is solved by means of simulated annealing and sequential least-squares quadratic programming from the SciPy library. The study considers the feature of solving the D-CRMP equation related to the errors in the reservoir liquid production forecast when previously shut-in well starting its production. The selection of the objective function and constraints, which are preferable when using the mentioned algorithms for D-CRMP history matching, is made. A method for choosing the best model is indicated when using an algorithm, which is dependent on pseudorandom number generator. The choice is made taking into account the predictive properties of the models. An approach to calculating confidence intervals based on the F-test is considered in detail. Evaluation of confidence intervals is caried out.