Kick & Loss Detection and Estimation using Distributed Models

Abstract

A kick in oil and gas drilling is an unwanted, unexpected leak of oil and gas from the reservoir into the well-bore. This can happen when the surrounding formation pore pressure exceeds the well-bore pressure. A loss of circulation is the leak of drilling mud from the well-bore into the surrounding formation which occurs if the well-bore pressure is sufficiently higher than the formation pressure. Both types of leaks can have severe consequences related to the safety of the drilling crew, the rig itself and the surrounding environment. Even if the risk of severe safety related incidents is minimal, leaks are associated with high purely economical costs, both related to down-time (non-productive time), or fracturing of the well which might affect the future revenue-generating capabilities of the well. Besides preventive actions to reduce the likelihood that a leak occurs, it is evident that the ability to attenuate a leak when they do happen is an important part of any drilling operation. Early kick & loss detection is an important factor in fast leak attenuation. Almost equally important is early kick & loss estimation where information about the leak, such as pore pressure, well pressure, formation permeability and inflow size, is estimated in real-time and used to guide the controller for faster leak attenuation. The well-bore can be several kilometers long and transient fluid flow effects are significant. The fluid flow is therefore often modeled by hyperbolic partial differential equations (PDEs). Previous results on kick and loss detection and estimation has mainly focused on using lumped ODE models (with some notable exceptions), where the infinite dimensional PDE model is approximated by a finite dimensional ordinary differential equation model. This thesis investigates the possibility of using distributed PDE models directly in kick and loss estimation schemes.