Multiphase flow meters are indispensable tools for achieving optimal operation and control of wells as these meters deliver real-time information about their performance. For example, multiphase flow meters located downhole can improve the production of multilateral and multizone wells by timely allocating the zone where a gas or water cone occurs. However, multiphase meters are either expensive, inaccurate, or cannot be used downhole due to the harsh conditions. An alternative that can be used to overcome these disadvantages is to use multiphase soft-sensors, i.e. to estimate holdups and flow rates from relatively cheap and reliable conventional meters, such as pressure and temperature measurements, and a dynamic model connecting these measurements with the unknown quantities. The aim of this paper is to demonstrate, via two simulation based case studies, some possibilities and limitations of such multiphase soft-sensors. In the first case study the question is adressed whether it is possible to use only downhole pressure and temperatures measurements to estimate in real-time the water, oil and gas flow rates in a well. This question is of practical importance as these measurements are relatively cheap and reliable. The second case addresses the question whether it is possible to allocate the gas cone in a well with multiple inflow points or zones. This question is relevant as the estimated flow rate and holdup profiles can be used to manipulate Inflow Control Valves in such a way that gas breakthrough is prevented. Using amongst others OLGA data as "real-life" data, an additional question addressed here is what the influence is of soft-sensor model error and measurement noise on the quality of the estimates. From the first case study it can be concluded that, due to bad observability, pressure and temperature measurements alone are not sufficient to accurately estimate in real-time well flow composition parameters in a practically relevant situation. The preliminary results discussed in the second case study indicate that a soft-sensing solution to the gas cone allocation problem may very well be feasible. Introduction Motivated by the ever growing discrepancy between demand for and availability of oil and gas and by the improvement and increased availability of downhole measurement and control equipment, the oil and gas industry has recently embraced the "smart wells" philosophy. The main idea of this philosophy can be stated as the improvement of current reservoir management by improving current reservoir and well monitoring and control practice. By doing so, one aims at a higher yield from a given reservoir, on the short-term and/or on the long-term, while simultaneously fulfilling constraints that are imposed out of environmental and (other) operational considerations. Here, the focus is on the improvement of current well monitoring practice. Well monitoring can be defined as real-time measuring or estimating well production performance parameters such as water, oil and gas flow rates. These can be delivered to an operator or a control system to allow for taking steps to improve current well production performance. In particular, monitoring devices located downhole can improve the production of multilateral or multizone wells by determining at which areas/zones of the well which fluids are entering. Even more specific, this knowledge allows for a better handling of gas or water breakthrough. See e.g. Leemhuis et al. (2007).

Real-time monitoring of downhole oil, gas and water flows in wells can significantly improve the production performance of these wells when this flow rate information is used to manipulate inflow control valves. An example of this is the allocation of a gas or water cone to its entrance point in a multilateral well, allowing to close down the individual well where the gas or water cone occurs, instead of closing down the complete well.Downhole monitoring of flows can be done via direct measurement. However, downhole multiphase metering is either expensive, inaccurate, or too difficult due to the harsh conditions. An alternative is to use softsensors. Softsensors estimate downhole holdups and flow rates from (relatively) cheap and reliable conventional downhole meters, such as pressure and temperature measurements, and a dynamic multiphase flow model connecting these measurements with the quantities of interest.Soft-sensing has already been investigated before for unilateral wells in Bloemen et al. (2004) and Leskens et al. (2008). In the second of these references, the simultaneous estimation of downhole oil, water and gas flows from downhole pressure and temperature measurements is considered. It is shown there that this estimation is badly conditioned (i.e. badly observable) and, thereby, not feasible in a practical situation. Using a similar approach and focussing on gas-lift wells, in Bloemen et al. (2004) it is suggested that soft-sensing with only downhole pressure and temperature measurements should work for the case that only a liquid and gas flow are estimated.In this paper, within the same soft-sensing framework as used in the mentioned two references, solutions are sought for soft-sensing of multilateral wells, both for the two-phase (gas and liquid) and three-phase (oil, water and gas) case.For that purpose, first, the question is addressed whether the unilateral two-phase case truly can be solved using only downhole pressure and temperature measurements. If so, the multilateral two-phase case is automatically solved with the corresponding soft-sensing solution simply consisting of a collection of unilateral two-phase sensors, one for each branch. It is shown that this solution is indeed feasible.After that, the three phase case is addressed. It is shown that for this case soft-sensing of multilateral wells is not possible, even when adding surface measurements and even though, as also shown here, it is possible for the unilateral well case when adding such measurements.

No abstract