Holistic flow and pressure control for underbalanced drilling operations
Abstract
This text is targeted towards engineers and researchers working with drilling
automation. Though the selected application domain is underbalanced drilling
(UBD) the results are applicable to managed pressure drilling (MPD) with or
without a two-phase fluid. The control goals are related to pressure and flow
management, or the hydraulic model, and not to areas such as geo-steering,
automatic pipe handling, weight on bit (WOB), or drill-string vibrations.
The main contributions are new results on multivariable control of flow and
pressure during UBD. Simple linear models and a more complex non-linear
model are used to predict the system behaviour during drilling operations. Both
standard model predictive control (MPC) and non-linear model predictive control
(NMPC) are used to generate control inputs. The benefits and challenges of both
approaches, together with the system performance, are discussed. The main
contributions are in the domain application, rather than in control methodology.
A main goal of the thesis is to look beyond the control of single sub-systems
to a more holistic perspective; where the objectives and constraints of the well
and the processing equipment at the surface are considered as one system. This
approach will ensure that one part of the system does not take actions which
are ill-suited for other parts of the system, and that the system may use more
degrees of freedom for control.
Manipulated variables range from pump liquid volume rate, choke opening,
separator pressure and level set-points, to rate of penetration (ROP); while controlled
variables include multiple well pressures, choke pressure, hole cleaning,
separator pressure and levels, separator gas flow, choke flow, and the choke operating
region. Constraints are included on well stability, production rate, pump
liquid rate, separator levels and pressures, choke flow rates, choke pressure,
pump pressure, hole cleaning, and gas production.
The control problem is configured and solved using industrial MPC software
already running in the field. Data is transferred using established industrial
standards for communication between process control systems and for model
exchange. The control system is implemented using a hierarchical approach,
with local stabilization of some of the underlying states. The necessary steps to
move towards a full industrial solution are briefly discussed towards the end of
the thesis.
An MPC solution is presented for control of both well conditions and the
topside separation system, using matrices of simple step response models. The
models are simple, fast, and easy to understand. The performance is shown to
be sufficient when evaluated for several scenarios, including connections and
drilling into new production zones. The simple models have limited validity,
and will have reduced performance away from the linearization point. However,
due to the slow time constants of the operation, it should often be possible to
keep the models up to date.
For the situations where linear models may not be good enough, an NMPC
controller which use a reduced drift-flux model (RDFM) is developed. While the
NMPC and the non-linear model are far more complex, harder to understand,
and the problem now takes much longer time to solve; it gives back in terms of
better predictive capabilities, it is valid over a much larger operating range, and
may require less model identification in the field.
We also hope that this thesis can serve as an introduction to UBD, with focus
on the control challenges and what can be achieved using automated control.