Politecnico di Milano - Dipartimento di Elettronica e Informazione

hammercoupleΜηχανική

22 Φεβ 2014 (πριν από 3 χρόνια και 4 μήνες)

182 εμφανίσεις


1

Politecnico di Milano
-

Dipartimento di Elettronica e Informazione

Piazza L. da Vinci, 32
-

20133
-

Milano (Italy)

claudio.maffezzoni@polimi.it


1.

State of the art

Combining technical and economic aspects in the power industry

In a scenario article by Hunt
and Shuttleworth, 1996, it was remarked that, in the
electricity market, “perhaps the need to separate the transport from the product would
not have arisen but for the realization that generation was no longer a material
monopoly”. Indeed, technology
-
pushe
d advances in gas turbine technology and the
possibility of burning natural gas to produce power made it economically feasible to
build and use smaller and cheaper generating units. Independent power producers
started to build their own plant (or plants) a
nd large consumers began to consider the
opportunity of meeting their energy needs according to a cost
-
based policy.


Since then (that is, since the 80’s), little progress has been made in Europe towards the
liberalisation of the electricity market, but s
ignificant changes have been introduced in
the power generation structure, because of the following concurrent facts:



conventional large power generation units have ceased to be the only economic way
to produce power;



independent producers have built a sig
nificant generative capacity;



cogeneration has been increasingly considered as a way to meet energy needs at a
lower cost;



environmental regulations have made traditional fuels (oil and coal) and traditional
power units increasingly difficult to use;



power

and heat have definitely become, at least in principle, products to be bought
and sold on a cost/profit basis.


In addition, governments (also in the European Union, by the 1996 Directive on
Electricity) are now ready to abandon protective policies in the

area of Electricity
Markets to promote competitiveness.


In this scenario it is clear that power generation must sustain itself by finding the most
economic way to produce electricity and possibly heat, in the case of Combined Heat &
Power (CHP) units. Ec
onomic optimisation of unit management and operation cannot
be based anymore on experience and practice only, since the continuously changing
market conditions (fuel price, environmental rules, customer demand, load variations,
tariffs and energy authority

regulations) may determine drastic modifications of the
optimal policies. Sophisticated Decision Support Systems (DSS) are needed both for
predictive scheduling on a medium
-
term horizon (from one month to a year) and for on
-
line scheduling on a short
-

or
very short
-
term horizon (from fractions of an hour to a
day). General solutions, valid for different plants and customer requirements, cannot be
found; on the other hand, the general guideline should always be that of minimizing
production costs while resp
ecting all the relevant constraints.


As a matter of fact, the use of complex software packages for optimised plant operation
is becoming more and more widespread, either as DSS for the assistance of plant
managers or as software tied to the DCS and intera
cting directly with the lower level

2

process control set
-
points. Examples of commercially available solutions include the
Boiler OP (Maines, Williams, & Levy, 1997; Levy, Cramer, Williams, & Sarunac,
1998) and GNOCIS

(Mayes, 1999; Noblett, Warriner, Logan,
& Faurot, 2000)
packages, targeted to the optimisation of coal
-
fired plant operation, as well as the
integrated Sienergy suite (Lausterer, 1999). Economic optimisation software for
cogeneration plants has been reported by Wolf (1994).


Models supporting de
sign and implementation of DSSs

Another key
-
point is that plant
-
wide optimisation directly interacts with low
-
level
control systems, and must be coherent with it, throughout the whole life of the plant,
from the initial conception, up to the commissioning
and to production, along with all
the changes and adjustment that are commonly carried out. An overall global system
model is then mandatory. Such a model is necessarily hybrid in nature.


Hybrid systems have been attracting more and more interest in the r
esearch world since
at least a decade. Actually, being at the border between time
-
based (continuous or
discrete time) models and discrete
-
event models, they pose new challenging problems in
the field of modelling, simulation and control. The transition fro
m laminar to turbulent
flow in a pipe, a phase change, the rising of the liquid level in a vessel above the weir,
are well
-
known examples of physical phenomena whose models can be simplified, for
specific purposes, by means of abstraction, assuming for the

system at hand a simpler
piecewise
-
continuous behaviour and introducing discontinuities, i.e. discrete
-
events, in
the model. In particular, parameters and time
-
scale abstractions can be distinguished.
The first occurs when small, often parasitic, dissipat
ion and storage parameters are
neglected, while the second occurs when a behaviour that occurs on a small time scale is
compressed to occur at a point in time. There are examples of such abstractions in
mechanics (stiction phenomena, collisions, etc.), in
chemical engineering (relief valves,
batch process dynamics and supervision, etc.), in the field of thermo
-
electrical power
generation and transmission (machine trips, plant coordination, start
-
up and shut
-
down
manoeuvres) and in other engineering fields.
In addition, there are hybrid behaviours
induced by the controller, such as the opening and closing of ON/OFF valves, the
switching of a control loop from manual to automatic operating mode, the action of a
controller at the end of each sampling interval.
For all such systems, a comprehensive
hybrid control system

(HCS) model is needed.

There are many approaches to hybrid systems modelling. Excluding those approaches
which take and enrich models from the discrete
-
event field with simple continuous
-
time
mode
ls, it is possible to derive HCS models from the classical models for continuous
systems, i.e.
differential algebraic equations

(DAEs), by introducing switching of the
system dynamics and jumps of the system’s variables. A general framework for HCS
that en
compasses several types of hybrid phenomena considered in the literature has
been proposed by Branicky et al.. Also, the object
-
oriented paradigm has been widely
exploited to describe HCS by different authors.

Among others, Barton and Pantelides have been
the first ones to introduce a rigorous
mathematical description of HCS with reference to the simulation problem of chemical
plants. Specific addressed problems are the change in equations and variable jumps.
Along that line, a simulation tool for industria
l purposes, gPROMS
(
general PROcess
Modelling System
)
, has been developed at Imperial College, London, to solve the
mentioned mathematical problems, that are not properly solved in most commercial
tools.


3


Unfortunately, the availability of good hybrid mod
eling and simulation tools may not be
enough to solve most practical problems in control and optimization of industrial plants.
Actually, models and tools must be integrated in the whole design process, so that also
their full exploitation and easy mainten
ance can be carried out.


Typical approaches to the control system design adopted in current practice are based
on a separation and sequentialization of the mechanical
-
thermo
-
hydraulic design and
control design. This leads also to the adoption of different

reference models for the plant
and machinery under control, which often do not agree with each other, being their
purposes quite different.


Among others, there are three main weak points in such a work procedure:

a)

neither systematic nor formal models
and methodologies are adopted for the control
design;

b)

there is no separation between what the controller should do, the so
-
called
control
law
, and how a correct implementation of such a control law can be obtained;

c)

there is no integration between the

mechanical and control system projects.


Point a) above means that a project is carried out with the help of intuition and
experience, which means that reusability of a previously designed and tested solution is
strongly limited. Also, errors and misunder
standings are captured very late in the
project, often in case of the final test or even installation of the whole plant.


Point b) means that control system is designed at a very low level, which of course
guarantees the feasibility of the control system,

especially with respect to single devices,
but makes it difficult to control a plant at the plant level. Also, the reusability of control
solutions is jeopardized, since they are coded in a specific language of a specific
controller.


Point c) means that
the various engineering activities involved in the design of a new
plant are not conceived as a whole: this implies that information on the machine is
spread on different models, and not necessarily kept aligned. On the contrary,
specification, design and
documentation should stem from a unique, unified system
model.


Beside the above points, presently there is also a strong push to lower down the design
time, in order to increase productivity of manufacturing companies, and to enhance the
flexibility of op
erating conditions, for the plant exploitation improvement.


All the above reasons push to adopt systematic metodologies to develop the entire
project, to adopt formal models, and, as for the control system, to separate
functions

from
code
.


A strong hel
p in conceiving suitable plant models comes from
object orientation
. The
object
-
oriented approach has been essentially proposed for software development, even
if it is based on quite general concepts such as modularity, abstraction, encapsulation
and inher
itance. In the late 1980, the object
-
oriented approach also started to become
popular in the field of modeling, especially with reference to physical system modeling.



4

One of its most valuable feature is that a modeling module corresponds to a concrete
ph
ysical object or component: the interfaces of the module with the rest of the system
correspond strictly to the physical ports by which the considered component exchanges
power or signals with the external world.


An abstraction hierarchy of the system bei
ng considered is realized by aggregating more
submodels and their connections within a composite (larger) model, which may in turn
be connected to other models.


Modularity, along with the standardization of interfaces, the aggregation of simpler
models wi
th abstraction, and the strict correspondence of model objects to real objects
are the basic mechanisms by which object
-
oriented modeling is superior in dealing with
complexity. To enhance the reuse of model
-
ware, object
-
oriented modeling
environments can
use inheritance to a certain extent, possibly by defining a specific
model type (or class) as a subclass of a more general one.


Control systems, like DCS, PLC networks or other types of hardware/software
architectures, should also be modeled in such a wa
y as to allow complex engineering
design to be effectively carried out, since many “details” contribute to system
correctness, besides to system performance. However, a comprehensive modeling may
not appear so natural to apply in this case, essentially bec
ause the real object (a control
system) may be considered from different (usually heterogeneous) points of view: a
functional point of view (a set of interacting algorithms related to control objectives), an
architectural point of view (a set of communicat
ing devices), or a software point of view
(looking at the different software layers and/or applications). Thus, it is not obvious
either what kind of decomposition/aggregation is the most opportune to split the overall
system into smaller modules, or which

are the behavioral aspects that should be
encapsulated by the individual objects.


A great help comes from the Object
-
Oriented Modeling Technique, and from the
subsequent, more elaborate Unified Modeling Language. Actually, besides notations
and other de
tails, such approaches propose important models (or views) of a system: a)
the state transition model, describing the (possibly dynamical) behavior and the
conceptual states of an object; b) the functional model, describing the flow and the
transformation
of data; c) and a model describing the main relationships among the
main abstract data (classes and objects). A main drawback of the object
-
oriented
approach is that the realized models are far from minimal. As a consequence, it is
important to reduce and
transform them, before generating code for real
-
time control, or
to adopt “traditional”, more efficient languages. This is why many known applications
of object
-
orientated modeling in control relate to high
-
level control functions.



2.

International Scientif
ic References


[
1
]

AA. VV. "Preparing for the 21st Century", Proc. of the 26th ASCE Annual Water
Resources, Planning & Management Conference, June 6
-
9 1999, Arizona, USA.

[
2
]

P.J. Antsakilis, A. Nerode, "Hybrid Control System
s: an introductory discussion to
the special issue", IEEE Trans. Automatic Control, Special Issue on Hybrid
Control Systems, vol. 43, No.4, 1998.


5

[
3
]

P.I. Barton, C.C. Pantelides. “Modeling of Combined Discrete/Continuous
Processes”,
AiChE J
ournal
,
vol.

40, pp. 966
-
979, (1994).

[
4
]

M. Branicky, V. Borkar, S. Mitter. “A unified framework for hybrid control”,
Proc. IEEE Conference on Decision and Control
, Buena Vista, Florida, pp.4228
-
4234, (1994).

[
5
]

H. Elmqvist
, S.E. Mattsson, M. Otter, "Modelica
-

An international effort to
design an object
-
oriented modelling language", Proc. Summer Comp. Simulation
Conf. '98, Reno, Nevada (USA), July 1998.

[
6
]

H. Elmqvist, F.E. Cellier, M. Otter. “Object Oriente
d Modelling of Hybrid
Systems”,
Proc. European Simulation Symposium
, Delft, (1993).

[
7
]

S. Engell. “Modelling and Analysis of Hybrid Systems”,
Proc. 2nd MathMod
,
Vienna, pp. 17
-
31, (1997).

[
8
]

gPROMS advanced user’s guide. Pr
ocess Systems Enterprise, (1998).

[
9
]

First International Conference on Computational Science and Engineering,
organized by SIAM in Sept. 2000.

[
10
]

P.M. Frank, "Analytical and Qualitative Model Based Fault Diagnosis
-

A Surve
y
and Some New Results", European J. of Control, Vol. 2, 1996, pp. 6
-
28.

[
11
]

Hunt, S., & Shuttleworth, G. (1996). Unlocking the Grid.
IEEE Spectrum 33
(7),
20
-
25.

[
12
]

Hoenig, P. and E. Welfonder (2000). Optimization of Short
-
Term Scheduling of
Power Plant Units in Large
-
Scale Cogeneration Systems. Proc. of the IFAC
Symposium on Power Plants and Power Systems Control, 26
-
29 Apr., Brussels,
Belgium.

[
13
]

R. Isermann and B. Freyermuth, ``Process fault diagnosis bas
ed on process model
knowledge
-

Part I: Principles for fault diagnosis with parameter estimation,''
ASME J. of Dynamic Systems, Measurement, and Control, vol. 113, pp. 620
--
626,
1991.

[
14
]

Lausterer, G.K. (1999). Sienergy


the Homogeneous S
olution for Power Plant
Management.

Proc. of the 14
th

IFAC World Congress,

397
-
402.

[
15
]

Levy, E., Cramer, D., Williams, S., & Sarunac, N. (1998). Application of Boiler
OP to Utility Boilers: Field

[
16
]

Maines, P., Williams, S
., & Levy, E. (1997). Application of "Boiler Op" for
Combustion Optimization at PEPCO.
Proceedings of the American Power
Conference 59
(1), 336
-
339.

[
17
]

Mayes, I.W. (1999). Generic NO
x

control intelligent system

(GNOCIS),
Colloquium Digest


IEE

65
, 9.

[
18
]

P.J. Mosterman. “An Overview on Hybrid Simulation Phenomena and Their
Support by Simulation Packages”,
Hybrid Systems: Computation and Control.
Lecture notes in computer science 1569
, Springer,
pp. 165
-
177, (1999).

[
19
]

Noblett, J.G., Warriner, G.H., Logan, S.R., & Faurot, M. (2000). Generic NO
x

Control Intelligent System


Results of Using Advanced Controls for Efficiency
Improvements and Emissions Reduction.
ISA POWID/EPRI Confer
ence 2000
, San
Antonio (TX), USA.

[
20
]

P. Putz, A. Elfving, "ESA's control development methodology for space AR
systems", In M. Jamshidi et al. (eds.): Robotics and Manufacturing: recent trends
in research, education and applications, vol.4,

ASME Press, 1992.

[
21
]

J.A. Stiver, P.J. Antsaklis, M.D. Lemmon. “A Logical DES Approach to the
Design of Hybrid Control Systems”, Department of Electrical Engineering,
University of Notre Dame, IN 46556, (1991).


6

[
22
]

Special

Issue "Computer Aided Control Engineering", IEEE Control Systems
Magazine, vol. 15, no. 2, 1995.

[
23
]

Special Session "Modelling of infinite
-
dimensional systems", organized by Prof.
Claudio Maffezzoni within 2rd Int. Symp. on Mathematical M
odelling, Feb 2
-
4,
2000, Vienna, 319
-
350.

[
24
]

Wolf, K. (1994). Economic Optimization Software Applied to JFK Airport
Cogeneration Plant.
Powergen Conference Proceedings 6D
(3/4), 153
-
162.



3.

The team


MAFFEZZONI CLAUDIO (full professor)

FERRA
RINI LUCA (associate professor)

LEVA ALBERTO (assistant professor)

PIRODDI LUIGI (assistant professor)

CASELLA FRANCESCO (assistant professor)

FRANCESCO SCHIAVO (PhD Student)



4.

References from Politecnico di Milano


[
2
]

M.L. Aime, F. Casella, C
. Maffezzoni: "Dynamic simulation of a co
-
generation
power plant based on gas
-
turbines", Proceedings of X ATI Conference on
"Tecnologie e sistemi energetici complessi
-

Sergio Stecco", Genova, June 2001,
pp. 371
-
383 (in Italian).

[
3
]

Aime M.L.
, Maffezzoni C. (2000). Modelling and simulation of combined lumped
and distributed systems. IMACS Journal of Mathematics and Computers In
Simulation. vol. 53, pp. 345
-
352.

[
4
]

V. Bolis, C. Maffezzoni e L. Ferrarini, "Synthesis of the overall b
oiler
-
turbine
control system by single loop auto
-
tuning technique", IFAC J. Control Engineering
Practice, Gran Bretagna, vol. 3, n° 6, pag 761
-
771, 1995.

[
5
]

P. Calabrese, F. Casella, F. Fiorani and C. Maffezzoni “System study of a
Geothermal P
ower Plant by Means of a Dynamic Simulator”, Proceedings of the
1999 European Control Conference ECC '99, Session CM
-
8, Karlsruhe, Germany,
31 Aug.
-

3 Sept. 1999.

[
6
]

E. Carpanzano, "A development methodology for hybrid control systems", PhD
T
hesis, Politecnico di Milano, 1999.

[
7
]

E. Carpanzano, L. Ferrarini, C. Maffezzoni, "Modular Testing of Logic Control
Functions with Matlab", 13th European Simulation Symposium, Marseilles,
France, October 18th
-
20th, 2001.

[
8
]

E. Ca
rpanzano e L. Ferrarini, "Simulation of a Process Control System affected by
Hybrid Phenomena", European Control Conference (ECC 2001), 4
-
7 September,
2001, Seminário de Vilar, Porto, Portugal, pag. 785
-
790.

[
9
]

E. Carpanzano e L. Ferrarini, "S
imulation of Hybrid Systems in Industrial Process
Control", Proc. 4th Int. Conf. On Automation of mixed Processes: Hybrid dynamic
systems, Dortmund, Sept 2000, pp. 29
-
34.

[
10
]

Carpanzano E., Ferrarini L., Maffezzoni C., Cataldo A. and Ceiner G
., "Testing
Industrial Distributed Control Systems with Hardware
-
in
-
the
-
Loop Simulators",
11th European Simulation Symposium and Exhibition, ESS’99, Simulation in
Industry, Erlangen
-
Nuremberg, Germania, 26
-
28 ottobre 1999, pag 574
-
578.


7

[
11
]

E.

Carpanzano, L. Ferrarini, C. Maffezzoni, "Simulation Environments for
Industrial Process Control", 11th European Simulation Symposium and Exhibition,
ESS’99, Simulation in Industry, Erlangen
-
Nuremberg, Germania, 26
-
28 ottobre
1999, pag. 443
-
450.

[
12
]

Carpanzano E. e L. Ferrarini, "Modular Modelling Of Hybrid Phenomena",
European Control Conference ECC’99, 31 Agosto
-

3 Settembre 1999, Karlsruhe,
Germany, Paper ID=F0654.

[
13
]

Carpanzano, L. Ferrarini e C. Maffezzoni, "An Object
-
Ori
ented Model for Hybrid
Control Systems", IEEE Int. Symp on Computer Aided Control System Design,
22
-
27 Agosto, 1999, Hawaii, pag. 132
-
137.

[
14
]

A Bottom
-
Up Methodology For Testing Complex Control Functions Of Process
And Power Plants
,
E. Carpan
zano, L. Ferrarini, C. Maffezzoni, accepted for 15th
IFAC World Congress, Barcelona, Spain, July 21
-
26, 2002.

[
15
]

F. Casella, C. Maffezzoni, L. Piroddi, F. Pretolani, "Minimising production costs
in generation and cogeneration plants", Control

Engineering Practice, 9(3),
pp.283
-
295, 2001.

[
16
]

Leva A., Bartolini A., Maffezzoni C. (1998). A process simulation environment
based on visual programming and dynamic decoupling. SIMULATION. vol. 71,
pp. 183
-
193.

[
17
]

A. Leva, C
. Maffezzoni, G. Benelli, Validation of Drum Boiler Models through
Complete Dynamic Tests, Control Engineering Practice, Vol. 7, No. 1, 1999, pp.
11
-
26.

[
18
]

C. Maffezzoni, L. Ferrarini e E. Carpanzano, "Object
-
Oriented Models For
Advanced Auto
mation Engineering", IFAC Journal Control Engineering Practice,
vol 7, 1999, pag. 957
-
968.

[
19
]

Maffezzoni C., Girelli R. (1998). MOSES: Modular Modelling of physical systems
in an object
-
oriented database. Mathematical Modelling Of Systems. v
ol. 4, pp.
121
-
147.

[
20
]

Maffezzoni C., Magnani G., "Practice and trends in control engineering", Lecture
notes in control and information sciences , Volume: 215 , pp.: 139
-
175,
ISBN/ISSN: 3
-
540
-
76060
-
1, (1996).

[
21
]

C. Maffezzoni,
G. Magnani, F. Casella, G. Bellani, S. Bandini, R. Mazzilli, M. De
Paola and S. Macchi, “Technological services management: an approach oriented
to the service quality evaluation”, Int’l Conference, organized by ANIE, GISI,
HYC, “The automation for the con
trol and the operation of public utility
networks”, Cagliari, Italy, 27
-
29 Sept. 1999, pp. 53
-
70. (in Italian).




5.

Description of suggested research


Plant
-
wide optimization and market
-
driven operation


When operating power plants in ever changing conditi
ons, criteria based on experience
and good
-
sense prove often far from being convenient from a technical and an
economic point of view. External conditions are dictated by costs and prices equilibria
in the energy market, by stringent environmental constra
ints and by different demands
from the Power System Operator who looks at ensuring safe technical conditions for the
system
. Moreover, different time scales must consistently be considered: medium term
production schedule (month
-
year) and short term schedu
le (down to fractions of hours),

8

with constraints relating both to peak values and to average values over different
periods.

Economic optimization is a DSS layer between process control and plant management:
it takes operation data and physical models fro
m the control level below and receives
technical
-
economical operating criteria from the plant management above. Scenarios,
events, contingencies and boundary conditions must be coherently updated and dealt
with at the different levels. To achieve this, it
is necessary that all the plant and stimuli
models employed, together with the time scales on which control and optimization must
occur, be consistent. Problems arise with plant
-
wide optimization over such time
-
scales
where plant dynamics and control are r
elevant, as this can lead to treat the control and
dynamic optimization problem in a context covering both technical and economic
issues, so that control design and economic objectives can confront each other in a
cooperative environment where the optimize
r is part of an overall dynamic system.

For complex power generation sets and various load profiles, the integrated approach
to control and economic optimization is quite necessary in the problem statement, but
needs appropriate decoupling/coordination me
thods for a practical problem solution:
the harmonization of these two conflicting requirements is a crucial research subject.

Moreover, operating power plants also on the basis of market issues can often mean
pushing them towards their technological limit
s under the currently accepted control
philosophies. This may in turn require reconsidering tools and methodologies for
assessing their performance. Since the preferred assessment tool appears to be
simulation,
it is then necessary to investigate the impac
t of market
-
driven operation at
all the relevant levels of the plant operation, so as to be able to include this impact in
simulation environments, where it may call for extending the validity of models to a
wider operating range than traditionally require
d
.

Environmental regulations have also an increasing importance in determining the power
(and heat) generation policies: research is needed to
provide good predicting models

of
emissions, effective tool of emissions control and reliable models for the pre
dictions of
concentrations, in order to avoid forced stopping of power generation with severe
economic penalties.



Self
-
maintaining and hybrid models


In the definition of the
overall dynamic system model
, the intrinsic modularity of the
plant should be e
xploited. However, an effective view of the system is the one where
each module includes the definition of both the control functions and a model for the
physical components, along with the physical interconnections among components, the
signal connection
among control functions, and suitable interaction between control and
physical components. In the above sense the model is called
hybrid
.


If we consider the typical components of a controlled plant, the following different
component types can be envisaged
:

-

physical (or process) components (e.g. pumps, tubes, heat exchangers, etc.);

-

control components (e.g. control devices);

-

instrumentation components (e.g. sensors and actuators).


Physical components are components of a plant which are able to perfor
m physical
transformations of materials and energy.

Control components are typically microprocessor
-
based devices which can execute
specified algorithms. To describe them, in general it is necessary to have
algorithms


9

(given in a suitable formalism) and
i
nterpretation/execution rules

describing how those
algorithms are connected with the rest of the world. With the adoption of suitable
interpretation rules and a specific target architecture the behaviour of control
components can be given only with algorit
hms at a very high
-
level of description, as it
is done in the current practice with the adoption of a
functional description
.

Finally, there are instrumentation components. They allow the connection between
physical and control components. Their descripti
on, in the general case, includes a
model describing the phenomena through which a command is actuated or a process
variable is sensed.

In the hybrid model all such components can belong to a unique module.


Modules can be described as
objects
, recalling t
hat the only part which is visible from
outside a module (an object) is the one declared in its interface, here constituted by
terminals
. Terminals are used to declare which quantities (or variables) can be
"exported" outside a module. Obviously, any other

information contained in a module
(equations, algorithms, variables, parameters,...) are internal data for the module.


Modules can be elementary or aggregate, if it results from the aggregation of elementary
or aggregate modules.

In particular, an eleme
ntary module can be given a behavioural model. Clearly, the
model will be given in suitable forms. For physical behaviour, equations (differential
and algebraic, possibly conditioned) can be used. For control behaviour, any
international standard can be ad
opted in principle.

More in general, also aggregate modules can be given one or more behaviour. A
behaviour is a suitable set of equations and algorithms which describes the whole
(hybrid in general) behaviour of an aggregate module and which is compatibl
e with the
module interface, i.e. with the module terminals. Clearly a behaviour coexists with the
hierarchical structure of an aggregate module (which defines also a model), but in a
simulation or in a formal analysis only one of the defined models can be

used,
according to the user specification. When using the behaviour, an aggregate module is
equivalent to a hybrid elementary module.

The necessity and importance of one or more behaviours, is evident. Suppose to have to
simulate for example the pressure
dynamics in a boiler, also the electric motors of by
-
pass valves of all the feed water pumps should be considered. However, this is not
feasible with the technology available now and that we can foresee to have in the next
few years. Consequently, we shoul
d consider the possibility to give an aggregate
module a
simplified model

(the
behaviour

field) to be used in different phases: detailed
simulation, simplified simulation, formal analysis, optimisation, testing and validation,
operator training.


Finally,
it is clear that it is not possible to adopt a unique modelling paradigm. On the
contrary, a multi
-
paradigm model can be adopted, where the appropriate description
model will be used according to the specific purpose (discrete
-
event models, hybrid
models,
rule
-
based models, conditional DAEs, fuzzy models, neural models, et.). This
item is a current research trend and practice as it shown by the Special Issue on
“Computer Automated Multi
-
Paradigm Modelling” to appear in the IEEE Transactions
on Control Syste
ms Technology in 2002
-
2003. Among others, a number of conceptual
and practical problems arise in the definition and maintenance of the overall dynamic,
hybrid, modular consistent models.


10

Building a consistent framework where aggregation is associated with
behaviour multi
-
modelling is an important research objective.


Experiment on a CHPP

Politecnico di Milano developed a number of significant experiences on the following
applications:



Conventional fossil
-
fired power plants



Combined cycle power plants



CHP pl
ants



Multi
-
unit with load distributions



Desalinization plants



Geothermal plants


For the proposed research, the research group is able to define and carry out suitable
experiments on the specific plant chosen as a demonstration test bed.



Summary

In a sum
mary, Politecnico di Milano can focus its research on the following items:



Investigate available simulation environments to check whether and up to what
extent they are suitable for treating the mentioned problems.



Develop suitable hybrid modular models, a
long with suitable manipulations
algorithms



Define typical optimization problems for power and CHP plants (or plants' pools)



Develop (some of) the necessary models and customize optimization algorithms for
special problems



Test the models' and optimizers'
performance in simulation, possibly comparing the
obtained results with real plant data if available.



Impact areas


Research results will have application in the following areas:



Supervisory control, optimization and management of power plants, CHPP and
of
generating plants pools, considering varying constraints and boundary conditions
(grid disturbances, market conditions, environmental restrictions,…).



Diagnostics of power plants and CHP plants, based on multi
-
paradigm modeling



Development of object
-
ori
ented model libraries for CHP and power plants, to be
used for off
-
line applications (simulation, design, control), and on
-
line applications
(optimization, diagnostics, analysis,…).