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

14 1. INTRODUCTION
boiler control concept, which includes automatic parameter identification of a process
model to be used for actuator and boiler controller design. This system must be capable
of allowing still smaller boiler geometry, meaning that the autonomy of the plant
must not be at the expense of water level and pressure performance. A complete control
system must also include fault detection and accommodation algorithms.
For control purposes, the boiler plant should be divided into modules representing e.g.
feed water system, burner and boiler. Each module has its own controller and the overall
controller has a hierarchy of control loops combined in a cascade configuration.
(This is of course a somewhat oversimplified picture as each module e.g. the burner
and oil system will contain both a loop for stabilising the oil pressure, heating the oil
and injecting the correct fuel amount to the furnace.) Such a configuration has many
advantages. Regarding control, inner loops on actuator systems are in general known
in the process industry to linearise valve gains and attenuate disturbances at the plant
input allowing an outer loop to treat it as a linear system. However, a much more transparent
advantage of such a strategy is the flexibility it allows in combining modules.
Of course such ability requires the aforementioned automatic parameter identification
to be implemented in order for outer loops to base their control actions on the correct
process model. When changes are made to the boiler plant, only the controller which is
assigned to the module that is affected needs to be updated. Simulink � and Real-Time
Workshop � supports automatic code generation for subsystems in a plant/controller
diagram.
It should then be possible to create a marine boiler control library containing models of
each process submodule to allow testing in a simulation environment. Apart from the
process models such library must contain controllers for each module. These controllers
should be general enough to allow simple configuration to adopt the controller to a
specific element within its class. E.g. boilers in the same family can use the same
control structure. The software tools for implementing the model and control library
could be e.g. MATLAB � /Simulink � in combination with Modelica [Modelica, 2007].
The latter supports object-oriented modelling of thermodynamical plants and can easily
interface with Simulink � . See e.g. [Casella and Leva, 2003; Eborn and Åström, 2000]
for Modelica examples of boiler modelling and power plant library design.
Regarding the control strategy, the scheme shown in Figure 1.3 supports the discussion
above and is the scheme used in this project. The inner loops around the actuators are
not shown in details in Figure 1.3. This scheme only applies to non-load sharing boiler
operation.
There are two new things to this figure; one is the green box including setpoint control
and burner decision control, the other is the external input. The external input covers
possible signals about load changes from the ship engine or cargo pumps. Such signals
can be used to improve the water level control in a feedforward and can be made
available. The setpoint optimisation is meant to use these signals together with possible
estimates of other disturbances acting on the boiler in order to calculate setpoints for
the water level, which are safe. Here safe means that the water level can be held within