Summary
This paper considers the use of extended Kalman filtering as a soft-sensing technique for gas lift wells. This technique is deployed for the estimation of dynamic variables that are not directly measured. Possible applications are the estimation of flow rates from surface and downhole pressure measurements or the estimation of parameters of a drift-flux model. By means of simulation examples, different configurations of sensor systems are analyzed. Finally, the estimation of drift-flux model parameters is demonstrated on real data from a laboratory setup.
Introduction
During the last 10 years, the industry has seen different downhole actuation technologies (commonly known as intelligent completions or under different trademarks) coming into existence. The goal of these technologies is ultimately to maximize the value of an asset by applying "right-time" optimization concepts borrowed from control engineering. Depending on the specific economics of the asset, this can be translated into more specific objectives such as speeding up of production, stabilization of unstable production, deferment of production of unwanted fluids, maximizing ultimate recovery, or a combination of some of the aforementioned short- and long-term objectives.
Control theory concepts of optimization by means of a feedback loop require means for determining the deviation between the actual response and the desired response of the system. In wells, this often boils down to some sort of multiphase flow measurement. Different accurate multiphase-measurement technologies have been matured during the last decade, and the industry seems to be crossing the chasm between the early-adopter and the early-follower stages. Often for control purposes, direct measurements with high absolute accuracy are not required, as long as the measurements give a good indication of the relative change in the property that needs to be optimized.
In different process industries, soft-sensing techniques were developed to determine variables where it is either impossible to directly measure the variables of interest or where it is economically not justifiable. In this paper, the concept of soft sensing is used; unmeasured dynamic variables (such as flow rates) are estimated from measured ones (i.e., pressures) by fitting a sufficiently accurate numerical model to the available measurements. We have looked at the gas lifted well application, where the lift gas rate may be controlled. Ideally this control would be based on directly measured multiphase flow rates, but in reality one often finds that this information is not available. Other measurements, such as surface and downhole pressure and temperature measurements, are more readily available and may be used for soft sensing.
The paper is organized in the following manner: first, the model of the gas lifted well is described; then, the soft-sensing concepts are explained; and, finally, different examples and configurations are shown in which this technology is applied for estimating multiphase flows.
Derivation of the Adjoint Drift Flux Equations for Multiphase Flow