ieee asia pacific conference and exhibition of the ieee-power engineering society on transmission...

Upload: saddam-hussain

Post on 06-Jul-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/18/2019 IEEE Asia Pacific Conference and Exhibition of the IEEE-Power Engineering Society on Transmission and Distributi

    1/6

    Transient Stability Preventive Control

    Using Critical Clearing Time Sensitivity

    J un ic h i

    Takaue

    K a z u y a

    Takahashi

    Tos h iya Oh ta ka Sh in ic h i Iwa m oto

    Waseda

    University, Tokyo.

    apanh

    Abst ract: Power systems have become larger and more

    complicated because of the increase of the electric power

    demand, and the analysis

    of

    a power system needs heavier

    computation. On the other hand, the social dependence on

    electric power is increasing,

    so

    the influence of ou tages in power

    systems is becoming more crucial. Therefore on-line preventive

    control is more important than ever. In this paper, we propose a

    generation rescheduling method for transient stability preventive

    control. First we sort all contingencies

    in

    the decreasing order

    of

    their severity. This process is done by sorting all contingencies

    in the increasing order

    of

    their critical clearing times (CCTs)

    computed using the dynamic extended equal-area criterion.

    Second we select several contingencies among them as objects

    of consideration, and compute generation rescheduling amounts

    using rotor angles

    of

    some machines at the appropriate time and

    the linearity relationship between CCTs and outputs of critical

    machines. To investigate the validity

    of

    the proposed method,

    simulations camed out using the 9-machine 20-bus system.

    Keywords: power system, transient stability, preventive control,

    critical clearing time, FEAC, contingency analysis

    1.

    Introduct ion

    Power systems have

    become

    larger and more

    complicated because of the increase o f he electric power

    demand, and the analysis

    of

    a power system n eeds heavier

    computation. On the other hand, the social dependence on

    electric power is increasing, so the influence of outages in

    power systems is becoming more crucial. Therefore

    on-line preventive control is more important than ever “ I .

    We must assess transient stability at the current

    operating condition using quantitative indices fast and

    accurately. However, it is very difficult to realize it

    because of the heavy burden of the analysis, so few

    examples to utilize on-line transient stability assessment

    exist. But many methods for

    on-line

    transient stability

    assessment such as the equal-area criterion

    [21

    and the

    transient energy function I” have been developed so far.

    Furthermore, the extended equal-area criterion

    [41

    that

    assesses transient stability using almost the

    same

    concept

    0-7803-7525-4/02/ 17.00 200 2

    IEEE.

    as the equal-area criterion w as developed recently.

    Incidentally, the relationship between Critical

    Clearing Times (denoted ‘CCT’ below) and outputs of

    critical machines has been found almost linear, and

    methods for transient stability assessment and transient

    stability preventive control have been developed [‘1[51.

    These methods compute the partial derivative

    of

    CCT

    with respect to the output of the most critical generator

    (denoted ‘CCT sensitivity’ below) numerically

    i.e.

    estimate the CCTs at the current operating point and

    another one using a method such as the transient energy

    function and draw a straight line connecting the two

    plotted points at the CCT-output plane, and find the

    intersection with the circuit breaker tripping time. And in

    [ 6 ] ,

    we proposed a method for computing the CCT

    sensitivity analytically using the parameters of the

    One-Machine Infmite Bus equivalent (‘OMIB’) obtained

    with the extended equal-area criterion. The aim of the

    method is to compute the CC T sensitivity faster than the

    numerical approach.

    In

    this paper, we propose a generation rescheduling

    method considering the first swing stability. First we

    screen contingencies using CCTs computed with the

    extended equal-area criterion, and estimate output limits

    for some specified contingencies using the computed

    CCTs and the CCT sensitivities. Second we determine the

    rescheduling amount using the estimated limits and the

    ratio of he magnitude of the rotor angles, and perform the

    screening for the operating condition after the

    rescheduling. If no unstable contingencies exist, we

    terminate the computation, and otherwise, we repeat the

    same

    process.

    To

    investigate the validity of the proposed

    method, simulations are carried out

    using

    the 9-machine

    20-bus system

    [’I.

    2 Outp u t l imi t Estimation using CCT sensitivity

    We

    explain the method for

    the

    estimation of the most

    critical machine (denoted ‘CG’ hereafter) output limit

    using the linear relationship between the C CT and the CG

    output. First we compute the CCT corresponding to the

    1518

  • 8/18/2019 IEEE Asia Pacific Conference and Exhibition of the IEEE-Power Engineering Society on Transmission and Distributi

    2/6

    current operating point using the Dynamic Extended

    Equal Area Criterion ('DEE AC '). Second we com pute the

    CCT sensitivity analytically using

    I ) .

    Defining

    P

    as

    mechanical output.

    I )

    represents the partial derivative

    of

    the CC T with respect to P,,,,(the sum of the mac hines

    belonging

    to

    the critical machine group). If the critical

    machines group used for the DEEAC has machine(s)

    different from CG, then we have to recalculate the OMIB

    parameters by letting the machines belong to the critical

    machine group.

    where

    Using the computed CCT and CCT sensitivity, we

    estimate the CCT-CG output graph and find the

    intersection with the straight line CCT=

    t,, t c ,

    is the

    circuit breaker tripping time), and the CG output at the

    intersection is the estimated CG output limit. For

    nomenclatures of this method is shown in Fig.1. The

    fundamentals of

    I )

    are describe d in detail in [ 6 ] .

    threshold l ine

    analytical direction

    't-JK;singurrent

    CT

    sensitivityperating point

    output

    stimated limit

    Fig.1 The method for the C G output limit estimation

    Require time for the analytical sensitivity

    computation is lesser than I13 of the numerical approach

    due to the unnecessary of the load flow computation and

    the CCT computation of the other operating point.

    However, considering the fact that the relationship

    between the CCT and the CG output is almost linear hut

    not perfectly, the approximation error of the CCT

    sensitivity computation itself exists, and the CCT

    computation using the DEEAC is not perfectly accurate.

    Thus it is possible that the estimated output limit is larger

    than the actual one. Therefore, a tuning

    is

    needed in order

    to prevent the harmful effects caused by the error.

    For

    example, set tC,+ktci k 2

    ),

    i.e. shift the thre shold line in

    Fig.1 somewhat upward.

    k

    defined

    as

    tuning coefficient

    and its value should be determined by the off-line trials.

    In

    this

    paper, we assume l.05.

    3. Transient stabil i ty preventive control

    In this paper, the stability scenario is that a

    three-phase short-circuit occurs at a generator bus at e

    subsequently cleared by tripping one line adjacent to the

    fault location at W C . G output value corresponding to

    CCT=t,, can be estimated by the method described in the

    previous chapter. We apply

    it

    to the transient stability

    preventive control that is performed using a generation

    rescheduling.

    3.1 Contingency screening

    Generally, power systems are very large systems, and

    it

    is

    impossible to assess the transient stability for

    all

    contingencies in detail. Therefore some of the

    contingencies have to be targeted as the particular severe

    ones (contingency screening). The procedure is as follows

    and the flow chart

    is

    shown in Fig.2. The numbers of

    0-

    @ in the following description on the procedure

    correspond to the one s in Fig.2.

    1519

  • 8/18/2019 IEEE Asia Pacific Conference and Exhibition of the IEEE-Power Engineering Society on Transmission and Distributi

    3/6

    I

    I J

    I

    I

    I

    Fig.2 The screening algorithm

    Compute the CCTs of all contingencies using the

    DEEAC, and sort all the contingencies in increasing

    order of the CCTs (contingency ranking). Further,

    identify CG corresponding to each contingency using

    the result of Critical Machine Ranking Method

    (‘CMR’) a+@+@.

    2 Check whether contingencies whose CCTs

    are

    smaller

    than

    tci

    (the group of these contingencies is defmed as

    A) are stable or not using the time-domain simulation

    program in the order of the contingency ranking. And

    go to step

    3

    if an unstable contingency is found @+@

    +a+@+D). If not, go to step 4 @+@+a+@

    +@+a).

    urthermore,

    unless A exist, go

    to

    step

    4

    @+@a)).

    3 The severest unstable contingency becomes ‘the main

    target’, and the corresponding CG is ‘the main CG‘. If

    an unstable contingency whose CG is different from the

    main CG is found in A, it becomes ‘the sub targetl’ and

    the corresponding CG is ‘the sub CGI’, and a stable

    contingency in

    A

    whose CG is different from the m ain

    CG and the

    sub

    CGI becomes ‘sub target2’ and the

    corresponding CG is ‘the sub CG2’. If the screening is

    not the fEst time and a m achine that was the main CG

    or

    sub CGI or sub

    CG2

    more than once up

    to

    the

    previous screening is neither the main CG nor the sub

    CG I, the machine becomes also the sub CG2 (This step

    is

    a-@

    s a w hole.). Next, compute the rescheduling

    amount.

    4 Check whether contingencies whose CCTs

    are

    larger

    than

    tct

    and smaller than

    t,,‘=tc,+O.05s

    (the group of

    these contingencies is defined as B) are stable or not. If

    an unstable contingency is found, it becomes the main

    target and the corresponding CG is the main CG. If

    contingencies whose

    CGs

    are different from the main

    CG are found in A, they become the sub target2 and

    corresponding CGs are the sub CG2. And whether A

    exists or not, if the screening is not the first and a

    machine that was either the main CG, sub CGI

    or

    sub

    CG2 more than once up to the previous screening is

    neither the main CG nor

    the

    sub CG I, then the machine

    becomes also the sub CG2

    @+@+@+@).

    If

    unstable contingencies are not found at all, the

    screening and rescheduling are stopped

    a+@+@.

    3.2 Generation rescheduling

    In this section, we propose a generation rescheduling

    method for transient stability preventive control. At first,

    we d efine the following notations The definitions of

    PO

    and NE are described later).

    mtc:

    main ta rge t

    sfcl: ub

    argetl

    stc2: sub target2

    M C G : main CG

    S C G l :

    sub

    CGl

    group

    SCGZ:

    sub CG2 group

    PO: group

    of

    machines belonging to PO

    NE: roup of machines belonging to NE

    The procedure of the generation rescheduling is as

    follows.

    1

    If contingencies belonging to

    the

    sub targetl exist,

    compute the sensitivity of its CCT with respect to the

    sub CGI output,

    and estimate the

    sub CGI output imit

    corresponding to CCT=f,t. On the basis of this result,

    determine the sub CG1 output change. Further, if

    contingencies belonging

    to

    the sub target2

    exist,

    the sub

    CG2

    output

    change is

    zero.

    The mathematical

    expressions are as follows.

    MI

    O

    V j E SCGZ

    (12)

    Pm,.-= p +

    w

    Vi

    SCGl

    (13)

    P,,... = P AP = P ,

    Q j

    E SCGZ 14)

    where CCTs,c,,Ds the CCT of the sub targetl computed

    using the DEEAC, and Pmi,nw(resp. P,j,nm) is the

    output of i th (resp. j-th) machine after the

    rescheduling.

    IS20

  • 8/18/2019 IEEE Asia Pacific Conference and Exhibition of the IEEE-Power Engineering Society on Transmission and Distributi

    4/6

    2 Compute the sensitivity of the CC T of the m ain target

    . with respect to the main CG

    output,

    and estimate the

    main CG output limit corresponding to CCT=f,,. On the

    basis of this result, determine the main CG output

    change. The m athematical expressions are as follows.

    MainCG

    p.,.._ = PA + M h

    Vh

    E MCG

    (16)

    where

    CCTm,c,D

    s the CCT of the main target, that is

    usually (when passing through

    I

    n Fig.2) computed

    using the DEEAC, but that has to be computed by the

    time-domain method when the main target is in B i.e.

    passing through ln Fig.2. And Pmh,., is the output

    of h-th machine after the rescheduling.

    3 Use 6 , t , ) , rotor angle at F I . ~ . btained by the

    Taylor series expansion, when identifying which

    machines belong to the critical machines group for the

    computation of the CCT of the main target. At first,

    classify all the machines except the main CG, the sub

    CG l and the sub CG2 into PO and NE. The

    PO

    is the

    group of the m achines whose

    rotor

    angles

    in

    the Center

    Of Angle ( 'CO N) reference frame at Ff are positive

    (the number of its elements is

    Np),

    and the NE is

    negative (the number of

    its

    elements is NnJ . Their

    rescheduling amounts are determined as follows.

    Urn cAc (17)

    kMCC

    SubCGI

    PO

    SubCG2 NE

    All the groups of the machines are compared in

    Table 1.

    Output

    change

    Decrease Decrease Decrease

    Constant

    Increase

    4. Simulation

    In order to investigate the validity of the proposed

    method, we demonstrate it using the 9-machine 20-bus

    system

    6]

    as shown in Fig.3 where all the generators are

    modeled as simplified models and all the loads are

    represented as constant impedance loads. We consider

    0

    -0 s

    the

    fault locations. The stability scenario

    is

    described in the beginning of the previous chapter. Set

    tc,=0.20s,

    and t,,'=tc,+0.05s=0.25s. Her e we do not chang e

    the output of generator 1 and 2, because they are operated

    in the rated values. Since generator 5 and 6, 8 and

    9

    are

    connected on the same bus, their behaviors are almost the

    same. Then we consider the generators belong to the sam e

    group.

    At first, the

    result

    of the contingency ranking using

    the DEEAC is shown in Table2 where 'CCT (TDS)'

    means the CCT computed by the time-domain simulation

    only to check the accuracy of the ranking and thus these

    need not to be computed for the actual operation.

    Comparing CCT(D EEA C) with CCT(T DS) to the ranking

    result, the ranking

    is

    proved

    to

    be accurate enough.

    As shown in Tablel, A is composed of

    0

    nd

    @,

    and

    B

    is of

    @, 0

    and

    0.

    he

    0

    egarded as

    the severest contingency is actually unstable on the basis

    of the result of the checking, and thus

    0

    s the main

    target

    and

    G5

    and G6 is the main CG. The

    0 hat

    is

    another contingency belonging

    to

    A and whose CG is

    1521

  • 8/18/2019 IEEE Asia Pacific Conference and Exhibition of the IEEE-Power Engineering Society on Transmission and Distributi

    5/6

    different from the main

    CG

    is actually unstable too, and

    thus

    @

    is the sub target

    I

    and

    G

    and

    G9

    is the sub CGI

    ,

    Su b target2 and sub CG2

    are

    not found in this case.

    The result of the screening is shown in Tahle3, and

    the generation rescheduling is performed using this result.

    Table2 Ranking result (initial condition)

    Table 3 Screening result (initial condition)

    Generator

    I

    2

    3

    4

    5

    6

    7

    8

    9

    0 t - h , [rad1

    Belonging group

    1.2972 Fixed

    0.0710 Fixed

    -0.0885 NE

    -0.0640 PO

    -0.1716

    Main CG

    -0.2 677 Main CG

    -0.2046 PO

    -0.

    330 Sub CGI

    -0.0970 Sub C G l

    Fault

    location

    apm8,9

    CCT (DEEAC) Check

    CG ategory

    Is1

    Where P is the sum of the output of G8 and

    G9.

    Then using the ratio of capacities, we determine the

    output changes of G and G9.

    0 1 0,190 3 Unstable 8,9(Su bCG I)

    AP9 =- c9 AJ

    - mm q-0 . 043)

    =

    -0.032

    C C, ' 555.55+1666.66

    Sub targetl

    The output of G8 and G9 after the rescheduling is

    computed using

    (1

    3 )

    (29)

    ns,,,er Pm8 AP, = 0.42 + -

    0.01 I ) =

    0.409

    Pmg n-

    P AP, = 1.25+ - 0 . 0 3 2 ) = 1.218

    Next, As is the case with sub targetl, using the CCT

    sensitivity of the a ( t h e main target) compute the output

    change

    Of

    G5 and G 6 using

    15),

    ap,,,,

    Where Pms bs the sum

    of

    output of G5 and

    G 6

    Therefore, the output changes of G5 and

    G6

    ( 3 1 )

    D5 =c

    1666 66

    x -0.088) =

    -0.066

    C,

    +C,

    AJ5 6= 555.55+1666.66

    -0.088) =

    -0.022

    p6=c 555.55

    C,+C, AJ5 6 =555.55+1666.66

    Therefore, the

    sum of

    the output changes of

    PO

    is

    computed using

    21).

    -

    33.3333 x

    { -

    0.088)+

    - 0.044))

    (34)

    3 . 3333+33 . 3333

    =-0 . 119

    1522

  • 8/18/2019 IEEE Asia Pacific Conference and Exhibition of the IEEE-Power Engineering Society on Transmission and Distributi

    6/6

    The sum

    of

    NE

    is

    computed using (22).

    M”*=

    -bpm