No abstract

A general dynamic model for any flow-related operation during well construction and interventions has been developed. The model is a basis for a new generation support tools and technologies needed in today's environment with advanced well designs, challenging drilling conditions and need for fast and reliable real-time decision support. The solution method uses a "divide and conquer" scheme, which computes the flow in each well segment separately, and then solves for the appropriate flow in the junctions. This simplifies greatly simulating complex flow networks, such as multilateral wells and jet subs. The flexibility allows incorporating additional pumps in the flow loop, as in Dual Gradient Systems. The model includes dynamic 2-D temperature calculations, covering the radial area affecting the well and assuming radial symmetry in the vicinity of the well. Other features: Flexible boundary conditions (which includes drilling, tripping) Non-Newtonian frictional pressure loss Transient well-reservoir interaction Slip between phases Advanced PVT relationship The reason for this new approach was to respond constructively to today's challenges (complex and difficult drilling conditions, need for reliable real time decision support). Our approach has been more flexibility, improved accuracy, reduced numerical diffusion and increased computational speed. The paper will present the basic model assumptions; the model architecture, and solution methods. Integration of the model into a real time system with links to real time databases and advanced visualization tools is currently ongoing. Thus the model may follow the whole work process through planning, training, execution, and post analysis. Examples of applications of the model so far will be presented, as well as visions for future applications. Introduction Alongside the advances in instrumentation and the more intelligent tools being developed today, there is a need for advanced numerical simulators that can bring all the technologies together to do intelligent drilling that will increase safety, reduce the costs, and make previously "impossible" fields "possible". The challenge in making the simulators lies in making the simulator kernel able to include all the important physical parameters; the important events; compute correct results; and fast enough to meet real time requirements. Examples of some of the recent improvements of the "events" that simulators are able to dynamically simulate are: control during underground blowout1; and hydrate formation2. In order to do this, the kernel needs to be able to change boundary conditions and connectivity during a single simulation. The simulator kernel presented here fulfills the needs of the present and many of the future challenges, as shown in the examples. The next chapters briefly explain the models used in the simulator, and the numerical methods used. The flexibility of the simulator is illustrated through a few examples, and the results of its use in real situation are presented.

No abstract

No abstract

There has quite recently been some discussion regarding the accuracy of gas kick slippage models used in transient flow models. This is not necessarily easy to give a correct answer on. However if pressure build up gradients for a migrating kick in a closed well is used for predicting the kick migration velocity, one can easily obtain a misinterpretation due to the fact that parts of the kick can become suspended in the mud. This was discussed in a paper presented two decades ago (Johnson et al., 1995). The purpose of the paper presented here is to show what kind of uncertainties that are involved in the transient flow modelling process and how that affects the pressure build up gradients. In this paper, a transient numerical model that can be used for simulating two-phase flow will be presented. It is based on the Drift flux model and the one dimensional advection upstream splitting method hybrid scheme (AUSMV). When using transient models for predicting the kick migration time, there are two sources of errors. The first is obvious in the sense that the gas slip parameters are uncertain. However, a less known potential source of error can be numerical diffusion and discretization errors caused by the numerical solver itself. The paper will show how the AUSMV scheme can be made less diffusive by using the slopelimiter concept. The paper will demonstrate that the effect of numerical diffusion can be quite substantial. This will be done by simulating pressure build up in a closed well varying both the slip parameters but also the degree of numerical diffusion. In addition, the simulated effect on the pressure build up when gas get trapped in the mud for lower gas concentrations will also be demonstrated. This paper aims at showing that the uncertainty caused by numerical diffusion can be of the same order as uncertainty in the slip parameters. It will also confirm the results reported in Johnson et al. (1995) by the use of a transient low diffusion flow model.