Vous êtes sur la page 1sur 36

Presented by: Ashish Khetan Indian Institute of Technology Guwahati Tutors: Prof. Ulrich Rde, H.

Kstler University of Erlangen-Nuremberg Germany

Indo-German Winter Academy 2007

Techniques of modeling
Introduction Object oriented modeling Component models Thermal stresses Analysis of fault events

Parallel ODE solvers for simulation


Introduction Richardson extrapolation method Parallel iteration method

Summary & conclusions

Power Plant Simulation

Schematic of a simplified fossil-fuel fired power plant

Power Plant Simulation

Schematic of simplified CCGT

Power Plant Simulation

Introduction

Steady state simulation


Transient Simulation

Thermodynamic design of water&steam cycle Design of components Part load behavior Pressure loss calculation Start up, shutdown behavior Thermal stress Massflow oscillations Design and study of control concepts Analysis of fault events

Power Plant Simulation

Introduction

Model structuring approach based on


Representation of plant components Interconnections between them

Physical ports
THT : Thermo-hydraulic terminal DHT : Distributed heat transfer terminal THHT : Thermo-hydraulic & heat transfer terminal HT : Heat transfer terminal MT : Mechanical terminal

Internal model description Software packages: APROS, LEGO, DYMOLA


6

Power Plant Simulation

Modular structure for heat exchanging system

Power Plant Simulation

Vertical heated circular tubes, risers, of evaporator Homogeneous model


Fundamental equations

Heat transfer calculations


Flow patterns Heat transfer regimes

Pressure loss calculation

Power Plant Simulation

Mass balance

Momentum balance

Power Plant Simulation

Component models

Boiler

Energy balance

Heat balance of tube wall

Power Plant Simulation

Component models

Boiler

10

Single phase liquid Bubbly flow Slug flow Annular flow Annular flow with entrainment Drop flow Single phase vapor

Power Plant Simulation

Component models

Boiler

11

Power Plant Simulation

Component models

Boiler

12

: additive friction factor for geometry elements : tube wall friction

Power Plant Simulation

Component models

Boiler

13

Control valve model

Governing equations h1 = h 2 1 = 2 w = f ( p 1 , p2 , h1 , y )

Power Plant Simulation

Component models

14

Governing equations po = p i + p p pp = fI (, q) h = fII (, q)

w(ho- hi) = H

Pump model

Power Plant Simulation

Component models

15

Governing equations
Flow equation, stodala law

Energy equation Power output

hi ho = ( hi hISO )

Pm = w (hi ho) m = Pm /

Power Plant Simulation

Component models

16

Need of analysis

Calculation of thermal stress values, with few assumptions, maximum value of tangential stress is

Thick walled components of steam generator and turbine are the limiting factors Spatial non-stationary temperature distribution Extreme positions Optimization of start up, shut-down or load changes Rapid operation implies more temperature excursions

Power Plant Simulation

17

Linear model, assuming thermal conductivity, density and the specific heat are independent of temperature space and time Radial heat conduction equation

Boundary condition

Large temperature excursions, non-linear model

Courtesy: G.K. Lausterer

Power Plant Simulation

Thermal stresses

18

Condensate pump failure in a feedwater system without buffers.

Courtesy: A. Butterlin, Erlangen

Where steam forms in the piping system and how far pressure decreases upstream of the feed pump ??

Power Plant Simulation

19

One dimensional heatable piping model Basic equations of the conservation laws for mass, momentum & energy with heat transfer equations Boundary points Simulation over time

Power Plant Simulation

Fault event analysis

20

Courtesy: A. Butterlin, Erlangen

Power Plant Simulation

Fault event analysis

21

Parallel processors Parallel methods for solving Initial-value problems for ordinary differential equations.
Explicit IVP methods (parallelism across the problem) Implicit IVP solvers (Linear algebra problem)

Parallelism across the ODE method

Methods with improved quality of the numerical solution Methods with reduced wall clock time per step
Richardson extrapolation method Parallel iteration method

Power Plant Simulation

22

A(h) be an approximation of A

Using Big O notation

Using h and h/t for some t

Solving the above two equations

Power Plant Simulation

Parallel ODE solvers

23

Increases order of accuracy of the given numerical approximation of true solution

Computing numerical approximations , i = 1,,r, where represents Romberg sequence

Power Plant Simulation

Parallel ODE solvers

24

can all be computed in parallel The are determined such that is more accurate than . Taking = 1, order of the extrapolation formula equals Q = q+r-1 Equations for determining , ,

Power Plant Simulation

Parallel ODE solvers

25

Given IVP,
Given generating method of order p Generating function with asymptotic expansion in powers of hs y(to+H,h) , numerical approximation y(to+H) , true solution y(to+H,h) identifies u() identifies hs Romberg sequence,

Power Plant Simulation

Parallel ODE solvers

26

Extrapolation formula

Explicit Richardson Euler method

Generating method, forward euler method

Yo = yo, Yj = Yj-i + hf(Yj-i), j = 1,2,....m y(to+H,h) = Ym , m = H/h

Power Plant Simulation

Parallel ODE solvers

27

System of equations Y = F(Y), F: Rdk Rdk


Y is the unknown function F is a nonlinear function

G is a free function with block diagonal jacobian matrix, the blocks of which are of dimension d Each set of d components of Yj is calculated independent of the other set of d components by Newton iteration.
Power Plant Simulation Parallel ODE solvers
28

Iteration method Yj - G(Yj) = F(Yj-1) - G(Yj-1), j= 1,2....

For the IVP RK4 method is Where

Slope is the weighted average


Parallel ODE solvers parallel iteration
29

Power Plant Simulation

Family of explicit RK method

Where

Power Plant Simulation

Parallel ODE solvers

parallel iteration

30

General form

Tabular form

Power Plant Simulation

Parallel ODE solvers

parallel iteration

31

Given IVP, General form of implicit RK methods, with k stages yn+1 = yn + hbof(yn) + hbTf(Y) , Y = yne + haf(yn) + hAf(Y)

It uses the average value of the slope at the different stages.


Parallel ODE solvers
32

e : column vector with dimension k with unit entries a, b : k dimensional vectors A : k by k matrix

Power Plant Simulation

Taking G(Y) = hDf(Y), where D is a diagonal matrix Iterative form of implicit RK method
Yj hDf(Yj) = yne + haf(yn) + h[A-D] f(Yj-1)
Initial approximation Yo - hBf(Yo) = yne + hCf(yne) B is an diagonal matrix and C is an arbitrary matrix

Power Plant Simulation

Parallel ODE solvers

33

Power plant can be simulated elegantly using the modelica script provided in the software packages which use the basic equations involving physical variables to model its components. These equations involve the partial derivatives, which are transformed into a much bigger set of ODEs. Parallel ODE solvers facilitate a way of solving these equations on parallel processors resulting in higher order of accuracy or reduced wall clock time per step.

Power Plant Simulation

34

Thermal power plant simulation and control, edited by Damian Flynn. Transient simulation in power plant engineering, transparencies of Siemens Power generation. Condensate pump failure in condensate preheater strings without a feedwater tank Dipl physics, A. Butterlin, Erlangen On-line thermal stress monitoring using mathematical models G. K. Lausterer Parallel ODE solvers P. J. van der Houwen & B. P. Sommeijer

Power Plant Simulation

35

Thank you
36

Vous aimerez peut-être aussi