THE STUDY OF THE PREDICTIVE ABILITY OF NUMERICAL AND ANALYTICAL MODELS (THE CASE OF MUTUAL WELL IMPACT EVALUATION)

Abstract

The main purpose of any model is to provide an opportunity to study the model object and the processes running in it for obtaining the predictive characteristics, among other reasons. In this connection, it is important to know, which mathematical models can help in analyzing and supporting oil deposit development, in particular, in assessing the mutual influence of production and injection wells. The characteristic features of mathematical modeling of field development include the oil deposits being located in natural formations that cannot be directly observed, as well as the complex filtration processes taking place in the formations due to the formation structure. Therefore, the mathematical modeling of development can be both complex and simple. On the one hand, it may use complex numerical hydrodynamic models, based on the understanding of spatial distribution of reservoir properties, which have an opportunity for detailed description of filtration processes. On the other hand, the modeling may use relatively simple analytical models, which have no need to specify the spatial distribution of properties; yet, the description of filtration processes is significantly simplified in comparison with hydrodynamics. Therefore, the practical value of the modeling result depends on the right approach to modeling. The task of estimating the mutual influence of wells requires the choice of numerical or analytical model to be based on understanding of the predictive ability of the models under consideration. Since such ability depends both on the ability to describe filtration processes in detail and on the need to take into account the spatial distribution of reservoir properties, it is initially impossible to conclude, which model has the best predictive ability. It becomes possible to reveal the level of predictive ability when considering the problem of mutual well impact assessment for synthetic models of oil deposits. This article presents the results of studies in the case of ten synthetic models. Numerical hydrodynamic models and analytical CRM models were set up for “actual” data of well operation. Using the retrospective test method, the authors have shown that the analytical models have a higher predictive power than the numerical models.