Abstract
Integrated simulation of reservoirs, wells, and surface facilities is becoming increasingly popular for modeling hydrocarbon production from deep offshore assets. Currently, there exist two common approaches for the integration. The first approach employs separate reservoir and facility simulators; whereas, the second approach combines the two within one reservoir simulation framework. Both approaches have advantages and drawbacks. For example, the first approach can be more accurate for the facility modeling, but overall it suffers from stability issues and long running times. On the other hand, the second approach is always numerically stable and typically provides better runtime performance, but requires additional inputs, e.g., Vertical Lift Performance (VLP) tables. Preparation of these additional inputs can be time consuming and often error-prone. Moreover, the VLP tables used in the second approach are typically constructed with the averaged values of "auxiliary" parameters, such as inlet temperature, water salinity, etc. This averaging can potentially lead to inaccuracies during simulation.
In this paper, we propose a new framework for integrated asset modeling which combines the benefits of the two approaches and hence significantly improves the efficiencies in both workflow construction and simulation accuracy. Our framework is based on the previously presented fully coupled network approach implemented as an in-house extension to a reservoir simulator. Here we extend the approach by introduction of an additional coupling step with a separate pipe flow (network) simulator. However, instead of using the pipe flow simulator to solve the network, it is used only to dynamically generate the VLP tables for the simulator's internal network module. Comparing to the previous fully coupled network approach, our new approach streamlines the simulation workflow by avoiding the necessity of the additional manually created network input. Furthermore this new approach also improves the modeling accuracy by using a generalization of the VLP description (e.g. with temperature as additional dimension) and by avoiding tables extrapolations. In this paper we discuss the new workflow and the new dynamic generalized VLP table construction in details.
