Modelling and dynamic co-simulation studies of oxy-fired power plant

Hannu Mikkonen1*, Mikko Jegoroff1, Jari lappalainen2 and Jouni Savolainen2

1
VTT Technical Research Centre of Finland Ltd, Jyväskylä, Finland
2
VTT Technical Research Centre of Finland Ltd, Espoo, Finland
*Hannu Mikkonen, +358 40 5953 909, [email protected]

Abstract

What kind of dynamic characteristics a new process concept will have? Is it easy to operate? Which
control strategy is the best? Suitable simulation tool will give answers to these kinds of questions.
Modelling and simulation software provide an environment for process modelling and accordingly, the
process phenomena can be simulated and design solutions can be evaluated. This paper focuses on
dynamic process simulation with one dimensional models. With simulations it is possible to carry out
process optimization, solve process bottlenecks and evaluate, for example, different alternatives to
integrate process areas before building the plant. The use of simulators for training is increasing
nowadays when the target is to make operators familiar with the process well before the
commissioning. There are different modelling environments for different needs. It depends on the case
and targets what kind of modelling would be the most fruitful, and finally, on financial issues what kind
of simulation, if any, will be implemented. The best cost-efficiency can be achieved when simulation
tools are used early in the engineering project, and the use is continued throughout the project, for
design, testing and analysis, for training and operation support as well. This article presents a study,
where Circulated Fluidized Bed Boiler (CFB), Air Separation Unit (ASU) and CO 2 Purification Unit
(CPU) were modelled dynamically using the Apros and Aspen Plus Dynamics programs. The dynamic
simulation of the whole CCS capable production chain ASU+boiler+CPU was achieved by combining
the tools, in other words, co-simulating the system.

Keywords: CFB, ASU, CPU, CCS, modelling, dynamic simulation, oxy-firing, co-simulation, Apros,
Aspen

Introduction

Depending on the development history of different simulation tools, there are plenty of differences in
solution methods, material properties, chemical reaction capabilities, ability to handle different phases
(solids, fluids, gases), and naturally, in the unit operation model libraries available for the users. Thus it
is quite common that a single tool does not fit all the requirements needed to fulfil the simulation study
targets. In this study Apros was used for boiler and turbine side, and Aspen Plus Dynamics for the air
distillation and CO2 purification side. Apros is very powerful tool for dynamical modelling of the power
plant and turbine islands, including solid, liquid and gaseous material streams, and it has also large
automation and electrical libraries. Aspen has very large chemical libraries as it has been developed
mainly for the needs of chemical industry. In this case the modelling tools were communicating
through the Matlab Simulink program.

This paper will present an overview of the novel CCS capable power plant concept, the modelling of
the process areas, the communication principles behind the co-simulation, and illustrative simulation
results of an operational transient. Finally, conclusions and future prospects will be given.
Process description

The boiler island design is based on Amec Foster Wheeler’s technology. The boiler island includes
feed systems for fuel, sand and limestone, primary and secondary oxidant streams, furnace, cyclone,
return leg, baghouse, heat recovery system to preheat the gaseous oxygen streams, flue gas
recirculating system, and different types of heat exchangers to transfer heat from flue gas and fluidized
solids into water, and on the other hand, from the water cycles to the gas streams. Flue gas circulation
is used only in the oxy firing mode. The circulated CO 2 rich flue gas is mixed with the O2 feed, thus
diluting the formed synthetic oxidant gas into a reasonable range of O2 content, to control the
combustion temperature and guarantee safe and economical operation of the boiler. The boiler model
gas side and the turbine island are illustrated in Figure 1.

Figure 1. The boiler and turbine island (Apros).

The furnace side heat transfer is separated in convection and radiation heat transfer and the user can
define the heat transfer coefficients or alternatively Apros will calculate them depending on the gas
and sand concentrations and flow velocities. During the load changes the bed solid material profile is
changing which causes characteristic dynamics for the heat transfer phenomena like in real boilers.
Also additional materials like iron and boiler masonry were included into the model to more accurately
achieve realistic boiler dynamic behavior. In the process side all the needed main equipment like
pumps, fans and valves were included in the simulation model.

From the control point of view there was a need to develop sliding pressure control for the main steam
pressure. Main steam temperature was controlled by using three spraying water valves. Feed water
amount was controlled by using a pump for pressure and a control valve for flow controlling according
to the boiler load. The boiler side control system included also a main load controller, which was
measuring the turbine power (MW) and controlling the fuel feeding. According to the fuel feed, the total
air/oxy demand was calculated and divided further for the primary and secondary air/oxy controllers.
Flue gas oxygen content was controlled by using separate controllers depending on which mode
(air/oxy) was in use.

The turbine island was separated in high, intermediate and low pressure part. In every part of the
turbine section efficiency and the taps were defined according to initial information acquired from the
turbine supplier. After the turbine, condenser level was controlled by a pump as well as condenser
cooling water amount. The condensate was then pumped to the low pressure preheating system
where it was preheated in four sections. There was also additional low pressure economizer where
part of condensate water was circulated and which was taking heat from the flue gas duct. The feed
water tank pressure was controlled by using a turbine tap valve. After the feed water pump and valve,
feed water was circulated partly via parallel high pressure economizer and high pressure preheating
system up to the main economizer and back to the boiler riser tubes.

The ASU plant operation is based on the cryogenic distillation. Distillation was made in a three column
system including low, intermediate and high pressure columns. At first, incoming air was compressed
in three sections. After compressing pressurised air was flowing to the HP and IP columns, where the
first distillation step was carried out. One part of the air flow was circulated via turbine which was
flashing air to lower pressure and at the same time cooling air to acquire colder conditions. Cryogenic
distillation is based on the evaporation temperature difference between oxygen and nitrogen. The
evaporation point of O2 is higher (-183°C) compared to N2 (-196°C) and this is why O2 is taken out from
the bottom of the column as liquid and N2 from the top of the column as gas. Finally in the first step
purified gases from the HP and IP columns were flowing to the LP column where final distillation was
made. For all distillation columns pressure and flow controllers were implemented in the model. The
pressurizing consumes a lot of energy and for this reason integration of waste heat from the
compressors to the boiler is necessary. The model included also the heat exchangers which are used
for inlet air cooling with the product flows (O2 and N2).

The CPU also consumes a lot of energy because of flue gas pressurizing. The pressurizing as well as
cooling is needed for incoming flue gas liquefaction. Also in this case waste heat energy has to be
integrated to the boiler part in order to improve the plant economy. The ASU and CPU models are
illustrated in Figure 2.

Figure 2. The Air Separation and CO2 Purification Units (Aspen Plus Dynamics).

In ASU and produced gaseous oxygen amount was controlled in two positions. Firstly, main oxygen
demand was calculated according to the boiler fuel load (as earlier described) and was given as a set
point for the ASU plant. Secondly, because of delays in oxygen production, O2 demand was given also
as a set point for gaseous oxygen buffer tank. This buffer tank has a capability to respond quickly
enough for the boiler load changes in most of the disturbances. The buffer tank consists of the
gaseous oxygen (GOX) header and liquid oxygen (LOX) tank. To enable sufficient pressure there
were two ways to handle the GOX pressure. When the load was decreasing, the GOX header
pressure was controlled by ventilating O2 out from the header (overpressure). During the load
increasing the pressure was maintained by evaporating LOX from the liquid oxygen tank. The
evaporating was carried out by using turbine tap steam in a condenser type preheater. Model diagram
of the oxygen buffer system is illustrated in the Figure 3.
Figure 3. Model diagram of the buffer system in the oxygen supply line (Apros).

The CPU plant was defined as a two train system. Two trains mean two similar compressor rows
which included four compressors each with same pressure proportion between the compressors. The
cooling was needed after every compressor to enable right gas conditions before the next pressurizing
step. After pressurizing, rest of water was removed from the flue gas before the purifying unit. Final
CO2 purification was made in separation tanks in 30 bar and -30°C based on flashing the gaseous
impurities out from the tank while CO2 liquid was taken out from the bottom of the tank. Liquid CO2
was then pressurized to supercritical condition (110 bar).

As mentioned above these two modelling tools Apros and Aspen were communicating through the
Matlab Simulink program. The interface is communicating in two directions so that one program is
calculating at the time. The results required for the co-simulation are given to the other program, which
uses the result data as boundary conditions in the interface of the models. The Matlab Simulink
application works as a master, which is taking care of the time synchronizing.

Connecting Apros and Aspen Plus Dynamics

Initial screening of dynamic linking techniques

Three techniques for linking Apros and Aspen Plus Dynamics were investigated:

Direct simulator-to-simulator OPC connection

OPC- and Excel-based connection
OPC- and Matlab/Simulink-based connection.

The direct OPC connection was realized with the Aspen Plus Dynamics’s OPC tools and the OPC
Server of Apros. It was noticed that connecting the simulators with this technique seemed to be
infeasible. The main reason for this was that both Apros and Aspen Plus Dynamics will simulate (i.e.
perform numerical integration of their model equations) independently of each other. In other words,
there is no guarantee that Apros’ simulation clock will run at the same pace as Aspen Plus Dynamics’
clock. This will pose problems especially when the two simulators run at a considerable speed
difference, since the faster simulator’s clock will run ahead of the slower one’s. Thus the values
transferred from simulator to another will not have any timely relevance to each other. In other words,
they are not synchronized. After this was noticed, this alternative was abandoned.
The second connection alternative was to use Excel as a mediator between Apros and Aspen Plus
Dynamics. The basic idea was to have Excel commanding both simulators to simulate for a certain
period, say 1s, and then wait until both simulators have done their job. Next data would be transferred
from simulator to another and the process repeated. This way the two simulators’ clocks would remain
synchronized. Aspen Plus Dynamics offered readymade connection to Excel and example Visual
Basic codes to achieve this. On the other hand, there was no direct way to connect Apros to Excel.
Thus, it was investigated whether we could utilize OPC to connect Excel and Apros. Three free Excel
OPC toolkits were tested, namely the OPC Office Link by Rensen Information Services Ltd.
(http://www.ris.co.nz/), Resolvica’s OPCEx (http://www.resolvica.com/) and Cogent’s OPC
DataHub (http://www.opcdatahub.com/Features/OPC_to_Excel.html). All three products offer easy
ways to get data from an OPC server to Excel sheets, but controlling a simulator through the link
proved to be difficult or even impossible. Such a control action, which was found to be missing, was for
example loading of simulation models. For this reason the Excel-based connection was abandoned
too.

The third connection alternative was to use Matlab Simulink as a mediator between Apros and Aspen
Plus Dynamics. The connection from Simulink to Aspen Plus Dynamics was realized with Aspentech’s
own interface and the connection from Simulink to Apros with OPC (using the “OPC for Matlab” -
product by Ipcos (http://www.ipcos.com/en/opc_for_matlab)).The entire linking system is depicted in
the following Figure 4.

Figure 4. Matlab Simulink-based communication.

This system was tested with “toy models” and with the ASU+Boiler model (ASU in Aspen Plus
Dynamics and boiler in Apros). Although some performance issues were raised, this way of linking
proved to be the most promising. A view of the connection in Simulink is shown in the Figure 5.
Figure 5. An Example of the Apros-Aspen link in Matlab Simulink.

Details of the connection

In Figure 5 we have one variable transferred from Apros to Aspen and eight variables transferred from
Aspen to Apros. The number of variables is configured to both Apros and Aspen Modeller blocks as
describe later. In addition to these, the Apros block takes in the current Simulink simulation time (T)
and the time step (DT). For the Aspen Modeller Block one first defines which model (.dynf) is to be
used. Next from the Input and Output tabs one defines which values are written to Aspen (Input) and
which are read from it (Output). For the Apros block configuration of the block is done in a Matlab
function simApros, which is in the m-file simApros.m. This code is called at every Simulink time step; it
performs the needed simulation operations. In the configuration phase the two variables readTag and
writeTag need to be edited by the user in order to accommodate the case at hand. Both of these are
vectors, each element of the vector containing a string. Each string defines one tag for the Apros OPC
connection.

Simulation of the system is done from Simulink. In Simulink the user defines the simulation time, in
seconds. Also, the user has a choice of numerical integration method, used by Simulink. As there is no
numerical integration to be done in Simulink, it is advisable to follow the recommendations of
AspenTech: “Within Simulink, the variable step ode45 (Dormand-Prince) integrator is recommended”.
Having defined these the running of the simulation is started by pressing the Play button of Simulink.
When the simulation is started the Aspen model is reset to time zero. For the Apros model this is done
already in the connection phase by connectToApros. Next the simulation cycle proceeds block-by-
block in Simulink. For example, if the cycle starts from the Aspen Modeller block the following steps
are taken:

1. Data is written to Aspen Plus Dynamics

2. Aspen Plus Dynamics is ordered to simulate for DT seconds
3. Data is read from Aspen Plus Dynamics
4. Data is written to Apros
5. Apros is ordered to simulate for DT seconds
6. Data is read from Apros
 The time to simulate during one cycle, DT, is given by Simulink’s solver. This six-step cycle is
repeated until the defined final time of simulations, T, is reached. During the simulation data can be
logged to files or to the screen from Aspen Plus Dynamics, Simulink or Apros. After the simulation is
finished the OPC connections are closed.

Conclusions on the connection

Above we presented results from the evaluation of techniques to connect the Apros and Aspen Plus
Dynamics simulators. The conclusion was that using Matlab/Simulink as the intermediator is
recommended. The tests showed that

data can be transferred from a simulator to another

simulator clocks can be kept in synch
models can be (re)loaded and the simulation experiment controlled.

The evaluation raised some future research issues. Firstly, it was noticed that when Aspen Plus
Dynamics is ordered to simulate for DT every simulation cycle, it takes a while to do some preparation
actions, before the simulation actually starts to run. This slows the system down. Thus it is
recommended that the computer in which the system is used, should be as powerful as possible. This
issue can also be alleviated by increasing the communication cycle (i.e. Simulink time step), as much
as the Apros and Aspen models and the dynamics of the system allow. Secondly, it was noticed that
the simulation models should be individually tested for the targeted simulation transients before
connecting them.

Results

The simulation study presented below conducts a dynamic load change from 100% to 70% and back
to 100% load. The simulation was carried out 15 seconds at the time after which Simulink applied the
main parameters for the other simulation tool and gave a permission to start simulation. First the boiler
load was kept steady for 18 minutes. After that fuel was ramped down 30% of the maximum load in 10
minutes. Then the load was kept steady for 54 minutes and after that ramped back to maximum in 13
minutes. In Figure 6 the main flows of the boiler are illustrated.
fuel ramped down 30% of maximum load in 10 minutes
then load kept steady for 54 minutes
ramped back to maximum in 13 minutes
100

80
Boiler load
was kept constant
for 18 minutes 60
kg/s 18
54
Fuel flow to boiler (kg/s)
40
O2 flow to boiler (kg/s)
10 13
54

20 Flue gas to CPU (kg/s)

0
0 25 50 75 100 125 150 175 200
Time/Minutes

Figure 6. The main flows of the boiler.

In Figure 7 the main steam parameters are illustrated during the load change. There was little
oscillation after reaching the minimum load which was a result of the boiler furnace solid material
profile.
700 temp
600
500 Main steam pressure
bar 400
(bar)
°C pressure
300 Main steam
kg/s
200 steam flows temperature (°C)
100 Main steam flow (kg/s)
0
0 25 50 75 100 125 150 175 200
Time/Minutes

Figure 7. The main steam values.

In Figure 8 turbine power is illustrated during the load change.

1000
800
600
MW Steam power to turbine
400 (MW)

200 Turbine production

(MW)
0
0 25 50 75 100 125 150 175 200
Time/Minutes

Figure 8. Turbine production.

In Figure 9 O2 concentration in flue gas, and primary and secondary oxidants are illustrated during the
load change. In this first generation concept of oxy firing, the O2-concentration in oxidants is close to
air firing values, typically between 21-28 mass-%. Flue gas oxygen was quite steady during the
change.
 0,3
0,25 Flue gas O2
concentration (mass-%)
0,2
mass-% 0,15 Primary recirculation gas
0,1 O2-concentration (mass-
%)
0,05
Secondary recirculation
0
gas O2-concentration
0 25 50 75 100 125 150 175 200 (mass-%)
Time/Minutes

Figure 9. O2-concentration in flue gas and primary and secondary oxidants.

In the following Figure 10 the main flows of the ASU are illustrated.

300
250
200
kg/s 150 Air flow to ASU kg/s
100
Produced O2 kg/s
50
Produced N2 kg/s
0
0
12,5
25
37,5
50
62,5
75
87,5
100

125

150

175
112,5

137,5

162,5

187,5

Time/Minutes

Figure 10. The main flows of the ASU.

Figure 11 shows oxygen and nitrogen concentration changes in ASU production during the load
change. At lower load level the produced oxygen concentration was decreasing and as a result of this
the efficiency of the whole system was decreasing.
 0,99
0,98
0,97
mass-% 0,96 O2 concentration mass-
0,95 %
0,94 Produced N2 mass-%
0,93
0

100

125

150

175
12,5
25
37,5
50
62,5
75
87,5

112,5

137,5

162,5

187,5
Time/Minutes

Figure 11. Oxygen and nitrogen concentration of the ASU product streams.

In Figure 12 the CPU main flows are illustrated.

100
80
60
kg/s Flue gas flow kg/s
40
20 Produced CO2 flow kg/s

0 Impurity flow kg/s

0

100

125

150

175
12,5
25
37,5
50
62,5
75
87,5

112,5

137,5

162,5

187,5

Time/Minutes

Figure 12. The main flows of the CPU.

Conclusions

This paper presented an overview of the novel CCS capable power plant concept, the modelling of the
process areas within two commercial dynamic simulation platforms, the communication principles
behind the co-simulation with the simulation tools, and illustrative simulation results of a load change.
The Matlab Simulink based method was found workable for co-simulation of the Apros and Aspen Plus
Dynamics simulators. The co-simulation approach proved to provide the benefits of both simulation
platforms in challenging operational transients. It is obvious that dynamic process simulation offers a
cost-effective and time saving way to develop new process concepts, evaluate the operability and
controllability with alternative control structures and thus, improve the maturity of the overall system in
early phase of the design.

The work with the full chain oxy combustion power plant concept continues. In near future, modelling
and simulation of the 2nd generation oxy combustion concept will be presented. The overall system
efficiency will be improved, for example with additional heat integration points between the ASU, boiler
and CPU plants, thus giving new challenges for plant operation and control.