Understanding the Performance and Application Window of Autonomous ICD

While it is understood that Inflow Control Devices (ICDs) can be an extremely valuable tool for reservoir management in certain applications, an engineering analysis is used to determine viability of ICDs for future developments. Typically, reservoir and production engineers use numerical reservoir simulation and/or steady state simulation to determine their applicability, evaluate various devices and configure their completions. Hence it is critical that the methods used to characterize the performance of the ICDs in these simulators are accurate. For passive ICDs, the characterization is fit for purpose; however for Autonomous Inflow Control Devices (AICDs), despite their recent rise in popularity, the performance characterization has been limited to mathematically derived correlations that are adjusted to initial data. By using these mathematically derived correlations rather than physics-based modeling, the characterization is unable to predict critical flow for compressible fluids and critical flow caused by cavitation in the case of pure substance flow, such as water. By neglecting the effects of the physical phenomenas such as compressibility and cavitation, the simulation will result in higher gas production than reality for gas coning control applications and higher water production in thermal applications where the operational condition is close to water saturation. Furthermore, fitting the correlations parameters for a wide range of fluid viscosities is complicated. This has led to several variants of correlation parameters that are dependent on specific ranges of viscosity. Hence, dynamic simulations can be complicated by the fact that the performance correlation factor has to be changed according to the fluid that is produced. In an attempt to remedy these limitations, a typical AICD, considered an industry standard, was modeled using a mechanistic approach to capture the physics of the process as closely as possible. Existing sets of test data, some of which were publically available, for fluids ranging from 0.011 to 500 cP, were used to tune the mechanistic model and test the ability of reproducing the data. The results have shown that the model has the ability to reproduce the flow loop test, effectively being able to predict compressible critical flow and performance of the device for a wide range of fluid viscosities. Furthermore, new experimental data was generated to test cavitation conditions that were included in a unified model. The unified model consolidates the expected operational window of the device, allowing accurate interpolation of non-tested conditions.