Due to geological reasons, there are randomly spread subsurface reservoirs with different pressurized fluids in them. These reservoirs are not easily detectable. Hence, when drilling a well one can penetrate such reservoir. Highly pressurized fluid mixes with the drilling fluid and a multiphase flow occurs in the wellbore annulus. Being highly pressurized, such flow inevitably creates hazardous conditions for drilling personal and environment. It is important to detect influxes into wellbores as early as possible in order to take urgent preventive measures. Therefore, real time monitoring of downhole flow conditions is necessary.
Different models exist for multiphase flow description. Excessive fidelity of one group of them does not allow for real time modeling due to large computational requirements. Noticeable inaccuracy of the other group of models does not allow for necessary flow characterization. Hence, the choice was made to employ drift-flux model to simulate an influx into a multiphase flow. Drift-flux model allows for quick flow simulation giving a reasonable accuracy.
To date, we have been able to simulate a mock transient gas influx into a single phase liquid flow. The form of drift-flux model we used is written for bubbly gas-liquid flow. We used time dependent boundary conditions to create a signature of influx. The results are presented on the figures – for a 500 m vertical pipe and 1000 seconds of simulation.
Drift-flux model contains three parameters which are highly dependent on flow conditions. Hence, the problems which have been solved with the help of drift-flux model are highly specific and applicable for narrow fields. We are aiming to describe a sufficiently wide range of possible multiphase flows with drift-flux model. For this purpose a fundamental research on drift-flux model parameters and mathematical behavior during an influx is necessary.
We plan to conduct a series of experiments as well as simulations on wide variety of flow conditions. Then we shall use multivariate analysis technics and, if necessary, machine learning algorithms to come up with reliable expressions for drift-flux model.