Optimisation of Marine Boilers using Model-based Multivariable ...

8 1. INTRODUCTION
Portfolio/plant
management
Units, plant
process control
Actuators: pumps, valves, etc.
materials/flow control
Economic optimisation
(static optimisation)
MIMO/decentralised control
(MPC, LQG, H∞, etc.)
SISO control
(PID)
Figure 1.2: Illustration of the process control hierarchy as employed in the industry. The different levels
seen from the bottom corresponds to task of increasing complexity and hence longer controller update times.
Further, the models used for controller design at the bottom two levels are dynamic whereas the models used
at the top level are often static.
cascade configuration. The models used in control designs at this level are dynamic.
The controllers at this level often have the purpose of linearising gains, reducing input
uncertainty and attenuating input disturbances.
The middle level controllers are the process controllers, e.g. the boiler pressure and
water level controllers. These controllers are often more complex than at the lower level
and the MIMO nature of the process is efficiently handled at this level. The update time
at this level is in the range of minutes or less, Ts ∼ min. Also the models used for the
controller designs are dynamic.
In relation to the marine boiler system the top level of the hierarchy could be seen as the
overall ship-wide control, if such control existed. This could be overall ship efficiency
optimisation. The update frequency of controllers at this level will usually be in the
range of hours Ts ∼ h. Furthermore, the models used at this level are often static. The
output of these controllers are future setpoints for the middle level controllers.
At both the top and middle levels it is often necessary to be able to handle constraints on
process outputs, e.g. water level, pressure or NOx concentration in the exhaust gas. In
this thesis we will work on the bottom and middle levels of the hierarchy. Especially we
will later address the model predictive controller (MPC) which is found at the middle
level.
Control methods
See e.g. [Åström and Hägglund, 2006; Franklin et al., 1994, 1998; Haugen, 1994;
Skogestad, 2003] for design and tuning rules for the PID controllers and feedforward
elements for SISO processes. It is possible to take the cross couplings in MIMO processes
into account while still allowing for a controller design based on SISO process