10149_06_paper

8/9/2019 10149_06_paper

1/30

S C H E D A E IN F O R M A T IC A E

VO LUM E 17/18 2009

Application of an Open Environment for Simulation

of Power Plant Unit Operation under Steady and Transient

Conditions

TOMASZ B ARSZCZ, P IOTR C ZOP

Depar tment of Robotics a nd Mechat ronics, AGH Un iversity of Science and Technology,

Mickiewicza 30, 30-059 Kra kw, P ola nd

e-mail: tba r szcz@agh .edu .pl

Abstract. The aim of the paper is to present a proposal and discuss an

application of an open environment for modeling of a power plant unit.

Such an environment is cal led the Virtual Power Plant (VPP) and is

based on a m odel created in th e Matla b/ Simulink environment. VPP

provides a framework for incorporating a broad variety of models,ranging from simple system models that run in real-time to detailed

models tha t will requ ire off-line mode to execute. The pa per presents t he

architecture of the VPP and briefly describes its main components. An

approach to implementation, including necessary simplification, sub-

models encapsulation and integration are discussed and i l lustrated by

schematics and equations. The paper includes a case study, where the

225 MW coa l fired un it is modeled.

Keywords:power plan t, st eam turbine, modeling, Mat lab, S imulink.

1. Introduction

Power plant simulators are developed for several reasons. They can be

divided int o follow ing ca tegories:

genera l purpose [1],

10.2478/v10149-010-0006-1Schedae Informaticae vol. 17/18 2009

Jagiellonian University Press, Krakw

by Jagiellonian University

8/9/2019 10149_06_paper

2/30

88

sta tic th erma l cycle calculat ions [2],

unders ta nding of a pow er genera tion process [3], w ha t if prediction [4],

virtua l prototyping of new softw ar e/ha rdw ar e plant components [5],

cont rol syst em optimisa tion [6,7],

improving of th erma l process performa nce [9],

environment a l concerns [9],

study ing a system beha viour out of the opera ting ran ge [10],

collecting diagn ostic relat ions among process feat ures of classified

components conditions for fault detection and isolation (FDI)

pur pose [11],

tra ining power pla nt opera tors thr ough pla ying/reconstr ucting

emergency scenar ios a nd ca se stud ies [10],

defining a nd va lidating safety opera tion procedures.

Virtual simulation of advanced systems plays an important role in

reducing t he t ime, cost a nd t echnica l risk of developing new solutions [12, 13].

The core part of a typical simulator is a model developed in one of

numerous simulation packages available on the market, such as PowerSim,

Aspen Dyn ., HYSYS, Ma ssba l, Mat lab/Simu link, Pr oTra x, Sinda /Fluint ,

Autodyna mics, MMS, APROS , gP ROMS, S IMODIS [14]. These packages ma y

have different functionality optionally using pre-developed power components

libraries [15, 16]. They are used for modeling of coal, gas, or combined power

units.

Ma tla b/Simu link is a very popular modeling envir onment [17]. It s

advantages and shortcomings have been analyzed considering the Virtual

P ower P lant (VPP ) applica tion scope and this pa cka ge has been fina lly chosen

as the core modeling environment. It provides an open and general-purpose

functionality. Therefore, many engineers and scientists are familiar with this

package. In addition, availability of auxiliary domain libraries (toolboxes and

blocksets) including the PowerSim blockset at reasonable costs is an

importa nt a dva nta ge. Matla b/Simulink package, in most a pplica tions, is used

to create simplified physical models suitable for purpose of control system

modeling [18, 19] however less applications refer to deep physical modeling.

Interesting example is the modeling of 677 MW coal- and gas-fired power

plant [20]. Another example is the Simulink model used for training in

advanced power plant process dynamics and control loop tuning [4]. It enables

simulation of differentiated operational scenarios including transient

operation. Literature also considers detailed physical models of steam circuit

components regarding fault detection algorithms and performance evaluation

[20]. A few Ma tla b/Simu link a pplica tions a re focused on a n electrical circuits

modeling including generator and power grid [e.g. 21]. Another interesting




8/9/2019 10149_06_paper

3/30

89

application is the validation process of a turbine regulator in respect of the

settings and hazard management [22]. An automaticcontroller is tested underdifferent operating conditions to detect disturbances in the system with the

use of a simula tor a tt a ched to the tur bine controller I/O signa ls [22]. In t his

case t he simula tor consist s of Mat lab/Simu link model a nd L a b-View controlled

Input /Out put ha rdw a re. All those a pplica tion ca se studies provide an

overview for Mat lab/Simulink strength s, disadva nta ges and implementa tion

process.

From th e modeling approa ch view point, th e Ma tla b/Simu link pa ckage

enables first-principle and data-driven model development. First-principle

modeling uses an understanding of the systems physics to derive a

mathematical representation. On the other hand, data-driven modeling uses

system test data to derive a mathematical representation. These two

approaches can be combined in application to modeling dynamic systems. The

advantage of the former approach is insight into the systems underlying

behavior and enables performance prediction, while the advantage of the

latter is a fast method for developing an accurate model and confidence

because it uses data from an actual system. The diff iculties of the former

approach are coefficients required to be determined, e.g. friction and flow

coefficient. The latter approach disadvantage is the need to handle multiple

data sets to cover range of system operation. Interesting comparison of

modeling based on physical principles and data driven model can be found

in [23].

For power generation applications authors in [24] proposed the Virtual

Power Plant (VPP), the innovative work environment intended for

reconstruction of a power plant unit functionality based on a model and a

recorded process data.

The paper is divided into three principal sections. The first section

introduces a simulator environment, while the second section presents the

st ru ctur e of th e core model developed in Ma tla b/S imulin k. The focus is on th e

structure and specific features, such as modularity and flexibility rather than

specific deta ils. The t hird section describes th e case st udy , w here th e 225 MW

coal fired unit was modeled. There are three subsections, discussing the

approach applied for the modeling of control systems, elements of the steam-

water cycle and the dynamic state. Since the subject is very broad, details of

chosen elements are presented. The fourth section presents results of

calibrat ion a nd va lidation of the model. This wa s part icularly diff icult t a sk, as

the t a sk could be only ba sed on t he recorded opera tional da ta .




8/9/2019 10149_06_paper

4/30

90

2. Virtual Power Plant environment

VPP, described in details in [25], provides a framework for integrating the

range of models, data management system, and visualization methods

including plug and play functionality. VPP consists of a few computers,

connected by a fast computer network (Fig. 1). The largest part of the system

is the database, which consists of two cooperating subsystems. The first one is

the typical DCS system. This approach allows to store the data in the same

way they are stored in a real plant . I t also al lows to present data in a user-

friendly w a y a nd t o intera ct wit h t he VP P from the level of mimic screens, like

plant opera tors use t o work. The second da ta base subsyst em is a specia lized,

fast database which is used to store data generated by modules of the VPP.This subsyst em is proprietar y, efficient da ta base engine, which can a lso store

dyna mic data (e.g. vibrat ion w a veforms).

Fig. 1.Structure of the Virtual P ower Plant

2. Submission of final version

The Central Bus (CB) is the central computer, which is the main data

exchange hub in the VPP. It provides common interface for all the modules,

which allows to develop each module independent from the others. It is

possible to exchange a module, or even to change the structure of the whole

system without changes in the software, but only in the configuration. The




8/9/2019 10149_06_paper

5/30

91

interface can exchange not only measurement and dynamic data, but also

events. Events are used to inform the selected modules about e.g. completionof a ta sk by a module and t hey a re also used to synchronize the whole system.

Other computers are used to run models of components of the power plant

unit. Number of computers depends on complexity and structure of the model.

Important part of the Virtual Power Plant are user interfaces, which are

closely connected with relevant databases. The first one is the user interface

native to the DCS system. It implements typical mimic screens of the unit

cont rol room. Thus, t he process can be monitored in th e sam e wa y a s it is done

by operators in their daily work. I t may be also used in the future to train

operators on the VPP. The other user interface access the data in the

specialized, fast database, delivering possibilities of advanced graphical data

analysis .

2.1. Structure of Virtual Power Plant model

The Virtual Power Plant Model (VPPM) uses a module-based and causal

a rchitecture ha ving th e possibility of referring to external librar ies w ritt en by

domain experts. These libraries can be developed in MATLAB-Simulink, but

can be also linked in t he form of compiled executa ble algorith ms or open codes

developed in C , C+ + , J a va, F ortr a n, or Ada . MATLAB -Simulink offers a

hierarchical object-oriented approach to communicate with OPC server using

the OPC Data Access Standard. I t allows acquiring live process data directly

into MATLAB-Simulink and writing simulation output to the OPC server. The

alternative is to exchange data through a temporary file using customized S-

function and read-write data converter. The simulation process is

synchronized with the system clock of the central bus (CB). The simulation

ca n be performed in on/off line mode dependen t on th e model complexity a nd

available computational power. In addition, it is feasible to generate and

compile C-code, being an equivalent of a developed model, accelerating

simulation process. VPPM implements major processes transformation of fuel

chemical energy into thermal energy, thermal energy into mechanical

rotational energy, and mechanical energy into electric energy (Fig. 2). Water-

steam properties are computed using steam tables based on the empirical

formulas which are the implementation of the IAPWS IF97 standard [26]. I t

provides accurate data for water, steam and mixtures of water and steam

from 01000 bar, and from 02000C. The current operating point of

w a ter/stea m mixtu re is propag a ted t hrough m odel blocks in th e form of a

state vector consisting of four state variables: temperature, pressure, mass

f low rat e , and entha lpy.




8/9/2019 10149_06_paper

6/30

92

Fig. 2.Power plant process flowchart

VPPM is split into Power and Vibration Module, which contain together

more then 25 libraries, 58 mdl-files, 3 s-functions and almost 100

initialization m-files. These modules can be run separately, exchanging

control and process data via the OPC server, and providing data

simultaneously with a real power generation process. The model must be

numerically stable and robust under steady and transient operation. A model

parameterization is a significantly time-consuming phase, because of large

amount of measurement data, machinery layouts, physical, and geometrical

parameters available as engineering specifications, manufacturer

documentation, control system manuals, turbine and auxiliary devices

manuals, and test reports [27]. The model files combine the libraries and

initializat ion files to creat e a VPP M a pplica tion of a specific power pla nt unit.

The libra ries cont a in t he component s models.

The auxiliary functions and procedures (e.g. load programs) are stored and

in the syst em folder. Simulink a llow s to crea te a libra ry of masked subsystem

blocks. This feature has been utilized in VPPM to create customized libraries

using the object-oriented data structures [12, 28]. Before starting simulation,

the initial conditions, load program, model parameters, controller settings,

and trip logic settings are uploaded into Matlab workspace (Fig. 3). PID

controllers and trip logic settings are required to initialize the governor

syst em w ith sa fety check limits, e.g. overspeed, min/ma x ra te limits for thespeed, load, and pressure. The initial conditions are required to run the model

at exactly specified operational conditions, e.g. at boiler start-up while the

rotor is stopped, before synchronization or at a given steady load. VPPM is

parameterized with geometrical and physical parameters including numerous

coefficients, constants and characteristics, e.g. valve opening characteristics.




8/9/2019 10149_06_paper

7/30

93

Fig. 3.Init ia l izat ion da ta required for the VPP M

Another key requirement to the model in Virtual Power Plant was the

possibility to keep several variants of model of a given component. It is

important, because development of the model is a process, where in the first

step independent partial models are developed. Those models may havearbitrary complexity, depending on the focus of research. Next, models are

interconnected to cover larger part of the power plant processes. A control

valve model is an example of increasing model complexity, where following

models were developed:

(i) ba sic version: sta tic experiment a l pressure-flow cha ra cteristic,

(ii) mediateversion: st a tic physical model including spring st if fness a nd

steam flow equa tion,

(iii) a dva nced version: dyn a mic physica l model including mediat e version

functional i ty + valve head inert ia .

Similar sets of variants are prepared for all modeled components. As a

result of the described process, to create the final unit model, the model of acomponent can be chosen from a list of available model versions. All such

models must have a common interface, but they will differ in model

parameters. The fundamental distinction is between steady-state and

transient models. Whereas the first one can be linearized, the latter is

inherently non-linear and thus, is much harder to develop. Therefore, each




8/9/2019 10149_06_paper

8/30

94

submodel also has its scope of application, i.e. valid range of input

parameters.

3. VPPM customization for simulation of 225 MW power unit

The model structure is presented and discussed using example of the

model customized for 225 MW power unit. The power unit is equipped with a

coal f ired wet bottom steam generator driving the three casing turbine with a

seven-stage boiler feedwater regeneration system (Fig. 4). Tab. 1 presents

basic para meters of the power unit under nominal loa d.

Tab. 1. B a sic power u nit pa ra meters referred to 225 MW operat ing conditions

The parameters of a turboset shaft line are given in Tab. 2. The rotor

critical frequencies are in the range between, i.e. 1300 1450, 17802230, and

27002880 rpm. The rotor is supported on 7 hydrodyn a mic bea rings, w hile the

total length of the shaft line is 29 meters. The shaft line model consists of 18

nodes. The rotor geometry is int roduced in det a ils in r eference [29] where t he

complete 160 nodes model of th e similar tur boset w a s described.

P ower unit

component

Properties Value

G enerat or P ower 225.6 MW

St a tor current 9919 A

P ower factor 0.85

B oiler OP -650 St eam production 650 t/h

Fresh stea m pressure (HP ) 13.8 MP a

Fresh steam tempera ture (HP ) 540C

Reheat ed stea m pressure (IP /LP ) 2.36 MP a

Reheat ed steam temperature

(IP /LP )

535C

Consu med energy 12301835 G J /h

Efficiency 93%

Cooling sys tem Cooling w a ter consu mpt ion 29 000 t/h

Cooling wa ter temperat ure 22 C




8/9/2019 10149_06_paper

9/30

95

Tab. 2. Turbine-generator properties

Fig. 4.P ower block functiona l scheme with t he sa mple para meters of steam-wa ter

mixture (EX stea m extra ction port, XW high pressure hea ter, XN low pressure

hea ter, CO condenser, PZ ma in pump)

3.1. Control system

VPPM implements the functional equivalence of a power unit control

system. The main part of the system is shown in Fig. 5. The complexfunctional group consisting of cascaded control blocks implement the control

concept of TU RB OTROL 6 (PROC ONTROL P 13 [30]), w hile ignores a uxiliary

contr oller a nd instr umenta tion details.

The basic controllers are speed controller, power rate controller and fresh

steam pressure controller. These controllers use signals from the fresh steam

pressure rate limiter, power rate limiter, speed rate limiter, temperature

Hea ding level High

pressure part

Intermediate

pressure part

Low

pressure

par t

Generator

Ma ss [kg] 7 800 16 300 49 982 42 650

Num ber of sect ions 12 11 4 N/A




8/9/2019 10149_06_paper

10/30

96

limiter after HP stage and nozzle diaphragm pressure limiter. The auxiliary

cont rollers implemented in th e Pow er Module ar e as follows: reheated steam temperat ure contr oller,

dru m level cont roller,

excita tion volta ge cont roller (AVR).

The automatic controller of Power Module governor, as explained hereinafter

by reference to Fig. 5, executes a pre-selected load program based on the

available process signals: (n) rotational speed, (no) steady state rotat ional

speed, (N) turbine electrical power, (T) steam and turbine casing temperature,

(Pex

) extr a ction pressures, (Po) outlet pressure, and (G) binary signal comma nd

latching the generator to a power grid. A thermal stress limiter operates on

the constrains such as the minimum boiler pressure Pmi n

, and casing

temperature T. The process constrains include the gradients of the rotational

speed, power and pressure. Control logic is provided operative in relation to

specific stages of the steam pressure and temperature build-up in sequential

order a nd selectively thr ough t he duct lines in prepa ra tion of tur bine la tching

a nd loa ding. This logic swit ches on/off valves of th e low a nd h igh-pressure

heaters banks. The control logic outputs a binary command signal on

synchronization line G, which latches a generator to power grid, and disables

the speed controller after a specific time. In addition, after the turbine has

been synchronized, steam is allowed to enter the low pressure heaters bank

through extraction outlets, and pipelines denoted as VII, VI, V, IV,

respectively to t he h eat ers XN3, XN4, XN5, cf. Fig. 4. Hea ters XN1, XN2

assembled in the condensers are in continuous operation with the condensers

CO1, and CO2. When the tur bine is loa ded at a given ra te, steam is a llowed t o

enter the high pressure heaters bank through extraction outlets, and

pipelines denoted as III, II, I respectively to the heaters XW3, XW2, XW1. The

regulatory control loops are coupled to the simulated sensors to receive

respective signals from the signal bus.

3.2. Power module

The power module (PM) consists of models of a steam turbine, low and high

pressure heaters, a boiler, a deaerator, condensers, mills, a control system, a

generator and a power grid. To formulate components models, unsteady

conservation equations for mass, energy and momentum have been used. In

order to implement the model in Simulink and to maintain the amount of

simulation time within the time available, some models were developed in

both advanced and simplified versions. All models components are connected

through ports enabling and propagating current steam parameters




8/9/2019 10149_06_paper

11/30

97

(tempera tur e, pressure, ent ha lpy), a nd/or ma ss/energy flow ra tes (Fig. 6).

Advanced models require the access to complete input-output vector, e.g. (T)temperature, (P) pressure, (H) enthalpy, while simplified models use only

mass flux (M) and temperature (T) variables. Other components models are

connected to the (M) mass flux port (raw coal, pulverized coal, steam, water),

(E) energy flux port (combustion, or mecha nical energ y), (N) rota tiona l speed,

a nd electr ical ports (U,I) (volta ge, curr ent).

Fig. 5.Functiona l scheme of the VP P M contr ol system




8/9/2019 10149_06_paper

12/30

98

Fig. 6.Steam-water circulation system implemented in Power Module includinga vaila ble sub-models ports

The str uctur e of th e P ower Module is shown in t he Fig. 6. Coal is conveyed

to a very fine powder in the pulverized fuel mills and mixed with preheated

air driven by the forced draught fan. The fuel controller through sequencer

regulates the supply of pulverized fuel to the boiler from four mills. This

process has a significant response time. A mill model involves first-order

dynamics of a conveyor and air-fuel mixture flow in the form of a transfer

function described in [18]. A mill control system is equipped with additional

sub-cont rollers a nd a mill sequencer, wh ich swit ches on/off th e par ticular

mills depending on the dema nded load . A hot a ir-fuel mixture is forced at high

pressure into the boiler, where it rapidly ignites [18]. The conservation of the

mass in the furnace assumes the ideal leak-tightened balancing of the mass

flow rate of the pulverized fuel, oil, air, slag, ash, and combustion gas [18].

This process is captured by a third order model simplified to a transfer

function.




8/9/2019 10149_06_paper

13/30

99

The heat energy is generated burning hard coal, and through convection

and radiation is absorbed by circulating water. The process of steamproduction consists of the pha ses:

(i) hea ting of w a ter in th e economizer,

(ii) evapora tion in th e dru m,

(iii) stea m overheat ing in th e superhea ter.

A reduced steam generator model implements dynamics of evaporation and

overheating while it ignores feedwater preparation process in the economizer.

The boiler model describes transformation of chemical energy into a heat

energy, which is passed t o wa ter f lows vertica lly up the tube-lined w a lls of the

boiler, and turns into steam. The investigations presented in [31, 32] show

that a drum model can be approximated with a second order model, or

optionally including steam a nd w a ter distribution in the drum a nd downcomer

pipe system in economiser with a fourth order model including momentum

balance [31] (Fig. 7). A set of controllers maintains a water level and steam

pressure [31]. It is necessary to establish the following assumptions for a

dru m model (sym bols fr om Fig.7):

(i) th e specific enth a lpy of stea m leavin g the boiler is equa l to th e va por

enthalpy h3= h

sa t(p),

(ii) th e pressure of feedw a ter is equa l to th e pressure of stea m,

(iii) hea t tra nsfer is domina ted by convection,

(iv) ideal heat tra nsfer between the feedw a ter inside the drum a nd the

surrounding meta l is a ssumed,

(v) the metal tempera ture is equa l to the satur a tion tempera ture of

wa ter for the pressure inside the drum .

Fig. 7.St ructure of the simplified drum m odel




8/9/2019 10149_06_paper

14/30

100

The steam passes to the superheater, w here its t empera ture a nd pressure

increase to around 12.7 MPa and 540C. The superheater model has beenderived using partial differential equations and then simplified to a series of

transfer functions representing the particular sections of a superheater [18].

The superheater model includes a water spraying process for purpose of steam

temperature control [18]. A schematic picture of a superheater section is

shown in Fig. 8.

Fig. 8.Superheater model

Fig. 9.Steam expansion curves




8/9/2019 10149_06_paper

15/30

101

The steam is piped to the high pressure part of the turbine. The steam

turbine contains stationary and rotating blades, grouped into three parts:high pressure (HP), intermediate pressure (IP), and low pressure (LP).

Stationary blades (nozzles) convert the potential energy of the steam

(temperature, pressure) into kinetic energy (velocity) and direct the flow onto

the rotating blades. The rotating blades convert the kinetic energy into forces,

caused by the pressure drop, which result in the rotation of the turbine shaft

[33]. Fresh steam is piped to the turbine inlet through trip throttle valves

(safety shut-off valves) driven by independent on-off hydraulic servomotor

wit h r ebound springs. These valves ar e located in t he steam supply line ahea d

of the governor valve. They are operated as the activating element for the

overspeed protection and serve as the manual throttle valve, which may be

used for testing and startup. They are designed to run fully open under

normal opera tion an d a re a ble to close very quickly. Next, the stea m is pa ssed

thr ough four pipes to the four contr ol va lves located in the H P turbine housing

and driven by four independent proportional hydraulic servomotors. They are

opened with oil pressure and closed with spring force. A physical model of a

hydraulic control system has been developed within previous research

programs [34, 35]. The whole model has moderately complex structure per a

single control valve including a second order servovalve model, a fourth order

linear double-side servomotor model, and a mechanical model of a massless

opening-closing valve head system. It is possible to simplify this system using

a second order linear transfer function per a single control valve. Valve

systems models are calibrated with the use of the measured data [27]. From

the control valves, the steam is passed to turbine through four nozzles. After

the H P part , the stea m is reheated in t he boiler. The reheated stea m is piped

to the valve chambers, where safety shut-off IP valves are located. Then,

steam is passed to the four IP contr ol valves connected t o the shaft cam dr iven

by a single servomotor described by the previously considered hydraulic

control system. All auxiliary on-off valves are modeled by approximated static

characteristics. The turbine is equipped with seven steam extraction ports,

from where steam is extracted for heating up the feedwater. I t is also

equipped wit h outlet/inlet ports t o/from th e superhea ter w here st eam is

passed for th e reheating process. St eam expansion curves ar e presented in t he

Fig. 9. The figure shows the steam parameters in the following locations: (1)

before shut-off valve, (3) exhaust I, (5) exhaust II, (10) before IP stage,

(11) exhaust III, (15) exhaust IV, (17) exhaust V, (21) before LP stage,

(22) exha ust VI&VII, (23) t urb ine out let.

Steam turbine dynamics can be modeled using the first principle model.

This method uses equat ions of the conservat ion of mass a nd energy in a steam

turbine [36] evaluating the work done by the fluid expanding in an

element a ry tu rbine volume [36]




8/9/2019 10149_06_paper

16/30

102

=

outin

mmd

dm&& ,

(1)

LQhmhmd

dHoutoutinin

&&&& += ,

(2)

d

dpVL =& .

(3)

The hea t exchange betw een t he stea m a nd the t urbines housing is described

by t he qua si-sta tic formu la [36]

( )TTkAQ = housing& . (4)In the low pressure part of the turbine, wet steam occurs, therefore the stateequations are defined in function of pressure and enthalpy, instead of

pressure and temperature, avoiding ambiguity. After rearranging, the

equa tions ha ve the form [36]

( ) ( )[ ]

Vp

v

vh

v

Qhhmh

vvmm

dt

dp

hp

outinin

p

outin

+

+

=

1

&&&&

,

(5)

( ) ( )[ ]

Vp

v

vh

v

Qhhmpvvmm

dt

dh

hp

outinin

h

outin

+

+

=

1

2&&&&

.

(6)

The volumes of stationary interblade channels are negligible. The static

characteristics provide relations between the input and output quantities of

an elementary turbine section based on the Fgl-Stodola equation adapted to

the transient conditions [36]. These characteristics allow to calculate steam

flow capacity, efficiency, power, and enthalpy drop in function of the

input/output pr essure, tur bine rota tiona l speed, a nd n umber of sta ges [36].

In second version of th e model, the t urbine dyna mics is a pproximated by a

simplified data-driven model. The model introduces time constants derived

from the mass conservation principle. The mass continuity equation

formula ted for a t urbine section is wr itt en as follow s [33]

outinmm

d

dV

d

dm&& ==

a nddt

dp

pdt

d

=

.

(7)




8/9/2019 10149_06_paper

17/30

103

Thus, t he stea m flow dela y is given by t he follow ing t ra nsfer fun ction [33]

outinmm && =

+11 ,

(8)

wh ere time consta nt is

pV

m

p

=

0

0

&

(9)

and where p0 an d

0m& are linearized values of the internal pressure and mass

flow rate, respectively. The delay constants have been applied for the inlet and

stea m chest d elay in th e order of 0.3 to 0.35 sec, crossover dela y in t he order of

0.4 to 0.6 sec and steam extraction pipeline in the order of 0.35 sec. The

additional delay is included in a reheater model. The enthalpy drop at eachsta ge of a condensat e multiple stage stea m tur bine (except th e last t wo sta ges

[38]), is assumed constant, based on the Saint-Venant and Wanzel equation

[38]. I f a steam turbine operates with constant rotational speed, the

peripheral speed is constant at each stage. Therefore, the speed coefficients

are assumed the same at all turbine stages [36] hence, the turbine section

efficiency does not depend on the transient conditions. In the simplified data-

driven steam turbine model the energy conservat ion equa tions are replaced by

static characteristics representing variable operating conditions at different

loa d a t par ticular tur bine sections where measur ements a re ava ilable.

( )in

mfT &=

a nd

( )in

mfp &= .

(10)

The steam para meters ar e approximated ba sed on th e experimenta l dat a (Fig.

10, 11). Steam extraction temperature, and pressure were measured under

differing load condit ions in t he r a nge of 160225 MW [27]. A stea m m a ss flow

rate was measured with the use of a calibrated orifice compliant to the ISO

standard [27]. The measurements were performed at the left and right inlet

pipelines to the t urbine, th e wa ter spra ying pipelines of th e fresh/rehea ted

stea m, a nd a fter t he feedw a ter pump before th e deaera tor (see Fig. 4) [27].




8/9/2019 10149_06_paper

18/30

104

Fig. 10.Approximated steam parameters at extraction ports (based on measurements)

The steam is condensing rapidly back into water, creating a near vacuum

inside the condenser. The condensers use a general heat exchange model

discussed in the next paragraph. The condensed water is then passed by a

feed pump th rough heat er banks, powered by a steam extra cted from th e high,

intermediate and low-pressure extractions, respectively. The temperature,pressure a nd enth a lpy of the condensa te/feedw a ter a re increased by a series of

low- and high-pressure heaters. The heaters with integral drain coolers are

the vertically arranged type with Utubes. A deaerator is a horizontal, direct

contact deaerating feedwater heater equipped with a storage tank. The

condensate is pumped to the deaerator, through XN12, XN3, XN4, XN5 low

pressure hea ters ba nk (Fig. 6). From t he deaera tor th e feedwa ter is pumped to

the steam generator through XW1, XW2, and XW3 high pressure heaters

bank. The feedwater heater drain system consists of drain removal path from

each heater. The normal drain flow path is cascaded to the next lower stage

heat er and t he alterna te pat h is diverted to the condenser.




8/9/2019 10149_06_paper

19/30

105

Fig. 11.Linearized relation betw een inlet steam m ass f low ra te an d extraction ma ss

f low rat es (based on mea surements)

Advanced and simplified heat exchanger models [10, 39] have been

developed for the purpose of modeling the feedwater heaters banks, deaeratorand condensers. A three-section advanced heater model is used when accurate

results are required. This heat exchanger model consists of three sections

including (A) desuperheating, (B) condensing, and (C) subcooling volumes,

respectively (Fig. 12). A steam circuit model assumes [10]:

(i) negligible heat exchange between the cavity an d the externa l

environment,

(ii) negligible heat accumulat ion in a wa ter, meta l housing of the

cavity, a nd pipelines,

(iii) negligible excha nges of energy and mas s, due to surfa ce

phenomena at the interface between the condensing and

subcooling areas,(iv) a ll heat -exchange a reas a re va riable a nd dependent on the

desuperhea ting , condensing a nd su bcooling volumes,

(v) uniformly distributed a nd consta nt pressure in the cavity equals

to the inlet steam pressure, uniform and averaged enthalpy

distribution inside each a rea (A, B, a nd C ) based on the boundary

conditions for each heater chamber,

(vi) negligible density va ria tions inside th e subcooling a rea .




8/9/2019 10149_06_paper

20/30

106

A feedwa ter circuit model a ssumes [10]:

(i) feedwa ter is in th e liquid sta te a nd in a subcooling condit ion,(ii) const a nt fluid pressure in th e tube-bundle equa ls to th e inlet

feedwa ter pressure,

(iii) uniform physica l propert ies of th e tube-bundle meta l,

(iv) negligible longit udina l hea t conduct ion in both th e pipe meta l a nd

the fluid.

Fig. 12.La yout of the th ree section hea ter model

The equations governing internal chambers energy and mass are formulated

as follows

56343443

34

+= Q

d

dpVQQ

d

dH&&& ,

(11)

67232332

23

+= Qd

dpVQQ

d

dH&&& ,

(12)

( )

=

dt

hhdm

dt

dH

hhdt

dm23

2323

23

231

,

(13)




8/9/2019 10149_06_paper

21/30

107

78121221

12

+= Qd

dpVQQ

d

dH&&& ,

(14)

( )

=

dt

hhdm

dt

dH

hhdt

dm12

1212

12

121

.

(15)

The equations below are formulated for the conservation of the energy in the

dra ining volumes, and they a re follow from the a ssumption of uniform density

of the wa ter neglecting va riat ion over t ime.

5634

56

5656

56

++= Q

d

dpVQQ

d

dH&&& ,

(16)

672367

676767

++= QddpVQQ

ddH

&&&,

(17)

7812

78

7878

78

++= Q

d

dpVQQ

d

dH&&& .

(18)

Introducing the weighted (average) steam-water properties, i .e. density,

temperature, constant volumes and constant pressure inside the heater

cavity, the heater model can be reduced to the two-section or a single chamber

heat exchanger model as presented in Fig. 13.

Fig. 13.La yout of the simplified heat er model

The P owerS im B lockset toolbox of Sim ulink ha s been used for modeling of

the generator submodel [25]. A ready to use electro-mechanical model of a

synchronous machine and a power grid were parameterized and integrated




8/9/2019 10149_06_paper

22/30

108

wit h other VP P M modules[25]. A breaker is opera ted t o connect th e genera tor

to the electric grid, when the generator acquires the proper operating point forsynchronization. A mechanical model of a turbine-generator considers

rotat iona l energy a nd a ngular form of vibra tions [40]. This model is a dequa te

under a ssumption of a sma ll deviat ion in r otat iona l speed an d consists of f ive

coupled rotor sections. The rotor section inertias H, damping factors D, and

rigidity coefficients K are assigned to generator, two LP rotor sections, IP

section, a nd H P section. The generat or is connected t o a n infinit e power grid.

3.2. Vibration module

The Vibrations Module (VM) is the separate part of the VPPM and models

the dynamic behavior of the shaft l ine. The steam turbine shaft rotates in the

casing on hydrodynamic bearings. The steam turbine coupled with generator

is a mechanical system of large number of degrees of freedom. Angular and

lateral vibration forms are mainly considered referring to generator driving

torque and rotor unbalance [40]. These two forms can be decoupled and

modeled separately neglecting the mutual influence of rotor eccentricity [40].

This assumption leads to angular vibration model used in the power module

(PM) and lateral vibration model used in the vibration module (VM). A torque

balance between rotor and generator is an excitation to the angular vibration

model, while th e eccent ricity a nd other sy nchronous/a syn chronous forces a re

an excitation to the lateral vibration model. Both models could be coupled, but

this w ould require huge increase in th e computa tion power.

The Vibrat ion Module is ma inly focused on t esting of different s cena rios of

early warning diagnostics. It is possible to convert the model response into a

hardware unit which generates vibration signals equivalent of real machinery

operation [41]. These simulated signals are the inputs to the vibration

monitoring syst em. They a llow to valida te its sensitivity a nd robustness under

different operating scenarios simulated by the model. Simulation scenarios

ma y conta in a lignment, ba lance, and incorrect cleara nce ma lfunctions.

A lateral vibration model considers the elastic shaft of the high

slenderness ratio consists of several rigid discs, mounted horizontally

embedded in hydrodynamic bearings arranged upon the supports [42, 43]. All

system mass is concentrated in the nodes. Linear, nonlinear infinitely short

and long bearing models have been developed [44]. Inertia, gyroscopic,

damping, stif fness forces, and excitation forces are associated with the n-th

node a s follows:

( ) nnxnnxnnxnnxn f uwwwKwDwGwM =++++ ,&&& . (19)




8/9/2019 10149_06_paper

23/30

109

A bearing model can be attached to an arbitrary node using the general expression

( )ww,&

f . The vibration module uses signal postprocessing methods [25] to plotvibration trends, cascade complex plots, orbit plots, etc.

4. Model validation

4.1. Static validation and calibration

After proper tun ing of the cont rol loops, th e dyna mic simulat ion model ca nbe operat ed in a st a ble steady sta te at t wo loa d points: 70% a nd 100% load .

The power unit model was calibrated statically using available performance

documentation, i.e. power unit energy balance. The calibration procedure of

th e pow er module consist s of th ree sta ges:

(i) module calibra tion,

(ii) group module calibra tion including local cont rol-loops calibra tion,

(iii) fina l pow er unit calibra tion.

The a va ilable documenta tion provides steam/wa ter m a ss flow ra tes,

temperatur es, and pressures for the opera ting r a nge wit hin 160225 MW wit h

a step of 10 MW. The model outputs are state variables and other calculated

or explicitly given values. The outputs from the model line up very closelywit h t he ava ilable data a t 70%and 100%(Ta b. 3).

Tab. 3. B a sic power u nit pa ra meters referred to 225 MW operat ing condit ions

Power unit component Error at 160 MW [%] Error at 225 MW [%]

Turbin e inlet 5.5 2

Turbin e inlet 0.19 0.08

Reheat er inlet 7.7 3.6

Reheat er inlet 0.33 0.1

Turbin e Out let 6.91 6.88

Tur bin e Out let 0.00031 0.6319XW1 feedwater after a

heater

2 2.6

XW3 feedwater after a

heater

0.2 1.7

XW3 st ea m to hea ter 2.7 34.5

XW3 stea m to hea ter 0.1 0

XW3 condensat e aft er a

heater

10 6.7




8/9/2019 10149_06_paper

24/30

110

One of the important aspects is model performance under continuous

operation, simultaneously to a power unit. A long-term test was prepared tovalidate the performance of power and vibration modules and available

simulation hardware, e.g. operational and cash memory, disk capacity, and

processor speed.

The load program is run continuously and it simulates daily or weekly

power plant operation under variable load conditions, e.g. combination of the

hot startup, steady state and coast down operation. The data produced by the

model is gathered and collected at the specified sampling frequency. The basic

model can be simulated in real time if i t is configured to exchange data

through read-write converter. A single workstation equipped with Intel

P entium 2.8 G Hz C P U an d 4 G B RAM operat ed under Microsoft Windows XP

P rofessiona l x64 edition, an d Ma tla b 7.2 (R2006a) wa s used for simula tion

pur pose. The follow ing s olver sett ings w ere a pplied: solver = ode23tb (st iff/TR-

B DF2), ma x step size = a uto, min step size = a uto, zero crossing contr ol =

disable all , relative tolerance = au to, absolute toleran ce = au to, da ta sa mpling

tim e = 60 [s].

4.2. Dynamic validation

The normal operating conditions were simulated to evaluate the

qualitative model accuracy compared to the reference power unit

measurements under closed-loop conditions. The steady state operation was

simulated with the use of a load program recorded during normal unit

operation. The simulation results were compared to the power unit response

at exactly the same reference signals, i.e. pressure and power. The power

plant operational data and simulation results were compared regarding the

steam mass flow and turbine demand load in Fig. 14. The small f luctuations

in the data are seen due to the assumed limitations of the power unit control

system model, and neglected or simplified components models. The valve

operation pattern is similar to real data as shown in Fig. 15; however, the

simulated valve opening waveforms do not follow adequately to the load

change. In Fig. 16, the st eam pressure simulat ed wa veform is compar ed to the

measurement data. The model resembles the steam pressure trend with

a ccepta ble accura cy. The reheated steam temperatur e compar ison is shown in

Fig. 16. The model reconstructs the trend, however exact temperature

va riat ions a re not included in the simulation results. In case of the feedwa ter

temperatu re, the t rend in t he simulat ion is t oo sensitive to load changes, most

likely it is an influence of simplified heater models used in this simulation




8/9/2019 10149_06_paper

25/30

111

(Fig. 17). In a heater model, the chambers volumes where fixed to simplify the

equations and increase numerical model efficiency.

4. Conclusions

The paper present s th e usa ge of th e Mat lab/Sim ulink packa ge to

implement the model of the power plant unit (VPPM), which is the basis for

the Virtual Power Plant (VPP). This environment facilitates virtual modeling

a pproa ch at component a nd syst em levels.

Fig. 14.St eam flow to the tu rbine (left) a nd t urbine power (right)

Fig. 15.High pressure (HP ) valves opening sequence (left) a nd feedwa ter ma ss flow

rat e af ter th e high pressure pumps sta ge (r ight)




8/9/2019 10149_06_paper

26/30

112

Fig. 16.Reheated steam temperat ure (left) and steam pressure at the turbine inlet

(right)

Fig. 17.Feedwat er temperat ure af ter t he high pressure heat ers (left) and turbine

outlet temperature (right)

The VPP can be run on standard workstations to play and simulate major

power plant processes in conditions close to the real time with accuracy

required for qua litat ive trend-based prediction and sensitivity a na lysis. VP P M

objectives are to decrease the uncertainty during preliminary power block

settings selection, better choice of the starting point in case of power block

optimization, identification of the most critical physical and geometry

parameters contributed to power block performance, and finally reproduction

of opera tional scena rios conta ined in th e measurement da ta .As presented in the previous sections the Virtual Power Plant Model

(VPPM) structure can be easily maintained and managed, due to introduction

of model variants, being model libraries updates. The available models

equa tions have been implemented an d integra ted in Ma tla b/Simulink. I t is

possible to use several modules creating a combination of simplified transfer

function models and extended advanced physics-based models within the




8/9/2019 10149_06_paper

27/30

113

single VPPM. Such an approach is important whenever model speed and its

flexibility are critical. It is possible to implement in the VPP several submodelversions to customize the model to specific needs of a modeling task, e.g.

transient or steady-state. A workshop has been organized together with the

involved power plan t specialists a nd a cademic sta ff to summar ize the sta tus of

the VPP and the VPPM development after the first phase of the project.

Developed VPP architecture was evaluated as fulfil l ing the project

requirements. However, i t is necessary to extend the VPPM to cover broader

operating range. The control system must involve more elements necessary for

good reproduction of a ll the syst em events.

The current project results can be divided into softw a re infra structur e an d

demonstra tion of the model for t he power plant unit. VPP project ha s included

development of powerful software infrastructure, predominantly for data

ha ndling, processing a nd presenta tion. Further r esea rch will be performed in

two directions: increase of the computational speed to achieve the real-time

operation and further development of the VPPM for better accuracy,

especially in tra nsient sta tes.

5. References

[1] Kim D.W., Youn C., Ch o B .H., Son G .; Development of a power plan t simul ationtool w it h GU I based on general pur pose design softw ar e, In ternat ional J ournal

of Control, Automation, and Systems, 13, 2005, pp. 493501.

[2] Thermoflow.Ava ila ble via : ht tp://w w w .th ermoflow.com/.

[3] Leva A., Ma ffezzoni C., B enelli G .; Val i dat ion of dr um boi l er models through

compl ete dynam ic tests,Control Engineering Practice,7, 1999, pp. 1126.

[4] PROVECTA . Ava ila ble via : h tt p://w w w .provecta .com.a u/.

[5] B ra ckenha mmer E., Schmidt E.; Dynam ic turbi ne sim ul ator for steam tu rbi ne

contr ol ler exami nati on. Availa ble via : ht tp://w w w.s iemens.com/.

[6] Ma ffezzoni C.; Boiler-turbine dynamics in power-plant control, Control

E ngin eering P ra ctice, 5(3), 1997, pp. 301312.

[7] Weng C.K ., Ra y A., Da i X.; Modeli ng of power pl ant d ynam ics and u ncert ain ties

for r obust contr ol synth esis, Applied Mathematical Modeling, 20, 1996, pp. 501

512.




8/9/2019 10149_06_paper

28/30

114

[8] Fra nke R., Rode M., Krger K.; On-l ine opt im izat ion of dr um boi l er star tu p,

P roceedings of the 3rd I nterna tiona l Modelica C onference, Linkping 2003, pp.

287296.

[9] La kshmira ju M., Cui K.; Num er ical i nvesti gati ons of pressur e loss reducti on in

a power pl ant stuck, Applied Ma th ema tica l Modeling, 31, 2007, pp. 19151933.

[10] Fly nn D.; Therm al power plants,Simulation and control, The Institution of

Electrical E ngineers, London 2000.

[11] nH a nce Technologies , Train i ng Simu lators. Ava ilable via:

ht tp: //w w w .nh a ncet ech.com/.

[12] va n P utt en H., Colonna P .; Dynam ic m odeli ng of steam power cycles. Part I

Simulation of a small simple Rankine cycle system, Applied ThermalE ngin eering, 27, 2007, pp. 467480.

[13] va n P utt en H., Colonna P .; Dynam ic m odeli ng of steam power cycles. Par t I I

Simulation of a small simple Rankine cycle system, Applied Thermal

E ngin eering, 27, 2007, pp. 25662582.

[14] G uima ra esa L.N., Oliveira N.S., B orges E.M.; Deri vation of a nin e vari able

model of a U-tube steam generator coupled with a three-element controller,

Applied Ma th ema tica l Modeling, 32, 2008, pp. 10271043.

[15] Elmega a rd B ., Houbak N.; Softwar e for t he sim ul ati on of power pl ant pr ocesses.

Par t A and B, E COS, 2002.

[16] Kim D.W., Youn C., Cho B .H., Son G .; Development of a power plan t simul ation

tool wi th GU I based on general pur pose design softwa re, In ternat ional J ournal

of Control, Automation, and Systems, 3, 2005, pp. 493501.

[17] MATLAB/ S IMU LI NK documenta t i on, Ma thw orks Inc., 2005.

[18] J a niczek R.S.; Power plant operation (original title in Polish: Eksploatacja

elektrowni parowych), Wydaw nictwa Na ukowo-Techniczne, Wa rsza wa 1992.

[19] Lu S .; Dynam ic modell in g and simu lat ion of power pl ant systems, P roceedings

of the I MEC H E , P a rt A, J ourna l of P ower an d En ergy, 213 (1), 1999, pp. 722.

[20] Cep k M.; Simu lat ion of the steam tur bine in M atlab environment, in 2nd

International Industr ial Simulation Conference Malaga, Spain , 2004, pp. 251

255.

[21] IE EE Subsy nchronous resonance working group, Second benchm ar k model for

comput er sim ul ati on of subsynchr onous resonance,IE EE Tran sactions on P ower

Apparatus and Systems, PAS104 (5), 1985, pp. 10571066.




8/9/2019 10149_06_paper

29/30

115

[22] B ra ckenha mmer E., Schmidt E.; Dynami c turbine sim ulator for steam tur bine

control ler exami nati on. Availa ble via : ht tp://w w w.s iemens.com/.

[23] Lu S., Hogg B .W., Dynam ic nonli near m odell in g of power pl ant by physical

pri nciples and n eur al network s, Interna tiona l J ourna l of Electr ical Power and

Energy Systems, 22, 2000, pp. 6778.

[24] B a rs zcz T.; Vi rt ual Power Pl ant i n cond iti on monitori ng of power generat ion

un i t, Proceedings of the 20th International Congress on Condition Monitoring

an d D iagnostic Engineering Man agement, F aro, P ortuga l , J une 1315, 2007.

[25] B a rszcz T., Czop P .; M ethodologies and appli cati ons of vir tual power pl ant

New envir onment for power plan t elements modeli ng,Insti tute of Sustainable

Technologies, Ra dom 2007.

[26] Holm gren M.; X Steam for M atlab.Avail a ble via : ht tp: //w w w .x-eng .com/.

[27] Test r eport, ENERGOPOMIAR, Poland, Gliwice 2005.

[28] Ma ffezzoni C., Ferra rini L., Ca rpan za no E.; Object orientated models for

advanced automat ion engin eeri ng, Control Engineering Practice, 7, 2000, pp.

957968.

[29] Kiciski J . , Drozdowski R. , Materny P . ; Nonlin ear m odel of vibrati ons in a

r otor-beari ngs system, J ourna l of Vibrat ion & Cont rol, 4 (5), 1998, pp. 519540.

[30] ABB PROCON TROL P contr ol and supervision w it h pr ocess opera tor stati onPOS30,User ma nua l , Germany, Ma nnheim 2006.

[31] Astrom K., B ell R.; Drum-boi l er dynami cs, Automa tica , 36, 2000, pp. 363378.

[32] Fr a nke R., Rode M., Kr ger K.; On-l ine opt im izat ion of dr um boi l er star tu p,

Proceedings of the 3rd

International Modelica Conference, Linkping 2003, pp.

287296.

[33] B oldea I.; Synchr onous generators, Taylor & Francis Group, New York 2006.

[34] B a rszcz T., Czop P .; System Ident i f icat ion and i ts l imi tat ion relat ing to

diagnosis of rotating machinery faults, Proceedings of the 10th

I E E E

International Conference Methods and Models in Automation and Robotics,

Poland, Midzyzdroje, 2004, pp. 10291034.

[35] B a rszcz T., Czop P ., U hl T.; Param etr ic approach t o fault detecti on an d i solation

in steam turbine control system, Proceedings of the 9th

IEEE Internat ional

Conference Methods and Models in Automation and Robotics, Poland,

Midzyzdroje, 2003, pp. 13571362.




8/9/2019 10149_06_paper

30/30

116

[36] U zunow M.; I nfl uence of discreti zation of a steam fl ow system on sim ul ati on of

tr an sient p r ocesses (in Polish), Ph.D. Thesis, University of Warsaw, Poland,

2001.

[37] Hobler H. ; Heat t ra nsfers and heat exchan gers (original title in Polish: Ruch

ciep a i wym ienn ik i), Wydaw nictwa Na ukowo-Techniczne, Wa rsza wa 1986.

[38] P erycz S. Steam and gas tur bin es (original title in Polish: Tur biny parowe i

gazowe), Zakad Narodowy im. Ossoliskich, Wydawnictwo Polskiej Akademii

Na uk, P ola nd, Wroca w 1992.

[39] G eorgia dis M., Ma cchiet to S.; Dynam ic modell in g and sim ul ati on of plate heat

exchan gers und er m il k fouli ng, Chemical Engineering Science, 55, 2000, pp.

16051619.

[40] Kr mer E.; Dynam ics of rotors and foundat ions, Springer-Verlag, Berlin 1993.

[41] B a rs zcz T., Ma ka M.; Appl ication of hardw are-in -the-loop for Vir tual Power

Plant, Proceedings of the 4th

International Congress on Technical Diagnostics,

Olsztyn, Sept 912, Poland , 2008.

[42] Muszyska A.; Rotordynamics, Taylor & Francis Group, Minden 2005.

[43] Ma hrenholtz O. et a l. ; Dynam ics of rotors stabil it y and system id ent if icati on,

Springer-Verlag, New York 1984.

[44] Frene J ., Nicolas D., Deguerce B ., B erthe D., G odet M.; Hydrodynamiclu bri cati on. Beari ngs and th ru st beari ngs, E lsevier, Amsterda m 1997.

Recei ved J anuar y 11, 2009

10149_06_paper

Documents