Вы находитесь на странице: 1из 8

IEEE Transactions on Power Systems, Vol. 10, No.

1, February 1995 43 3

THE HARMONIC DOMAIN. A FRAME OF REFERENCE FOR POWER


SYSTEM HARMONIC ANALYSIS.

J. Amllaga A. Medina M.L.V. Lisboa M. A. Cavia P. SAnchez


FIEEE Non-Member Non-Member Non-Member Non-Member

University of Canterbury, New Zealand University of Cantabria, Spain

Abstract - The Harmonic Domain is a general frame of ref* The above Considerations lead to a frame of reference which
rence for power system analysis in the steady state which mo- must be multi-phase and multi-harmonic, capable of represen-
dels the coupling between phases and between harmonics. In ting the mutual effects between phases and the cross-coupling
this frame of reference the non-linear components, converted between the harmonic freqnenaes; it must also be able to pro-
into harmonic Norton equivalents. are combined with the rest cess simultaneously the presence of voltage and current sour-
of the system and solved iteratively by the Newton-Raphson ces of diffenxt frequencies.We call such frame of reference the
technique.This paper d e s c r i h the structure of the new domain Harmonic Domain (HD) and the purpose of this paper is to
and ilustrates its potential applications in a small power sys- describe its framework and application.
tem with multiple non-linearities.
Kewords - Harmonics. Simulation. Power Systems Analysis INITIALSTEADY STATE SPECIFICATION

INTRODUCI'ION The steady state operation of a power system is defined by


a load flow program, which provides the fundamental voltage
In conventional power system steady state models the only phasors at all the specified nodes.However, for a precise defi-
active power sources are the generators, normally specified as nition of the base operating point of the non-linear compo-
voltages sources of a single frequency at their terminal busbar. nents, the load flow must be performed in the three-phase
Single phase (symmetrical) studies are considered sufficient !s frame of reference.
analyse the fundamental frequency behaviour of the system.
The use of direct phase coordinates to represent mutual
The increasing presence of distorting elements in power couplings between phases and between harmonics avoids the
systems requires accurate prediction of the harmonic content. need for complex transformations. The advantages of direct
Such prediction can not be achieved by conventional power phases coordinates have been demonstrated with practical
system software. Two important considerations in the deve examples in refs [3], [4] and [12].
lopment of a suitable algorithm, for harmonic modelling, are
the frame of reference and the nature of the sources. The nodal information derived from the three-phase load
flow needs to be modified for use in the Harmonic Domain as
Experience with field tests seems to indicate that the har- follows:
monic content in power transmission systems is always asy-
metrical, and its prediction must therefore be canied out in the At linear load nodes, the load connected to phase m of a bus-
three phase frame of reference. bar k is calculated in admittance form as:

Regarding the nature of the sources, a distinction needs to


be made between the fundamental and harmonics frequencies.
For the fundamental frequency the specification of the genera-
tors as voltages sources remains valid. At harmonic frequen-
cies, howewer, the non linear components -including the sa-
lient pole generators- act as current harmonic sources as well
and at a harmonic of order h
as harmonic sinks.

94 SM 604-9 PWRS A paper recommended and approved


by the IEEE Power System Engineering Committee of the
IEEE Power Engineering Society f o r presentation at
the IFXE/PES 1994 Summer Meeting, San Francisco, CA, the h in the denominator of equation (2) is an approximation
July 24 - 28, 1994. Manuscript submitted July 26, based on the fact that r c< x in a transmission network.
1993; made available f o r printing May 6, 1994.
At generator nodes and fundamental frequency the specified
terminal voltage and complex power are converted into an in-
ternal emf behind synchronous reactance.
0885-8950/95/$04.00 0 1994 IEEE
434
At each harmonic frequency the generator is normally repre- +m
sented by an appropriateimpedance. However in the presence f'(Xb)= Cidw (74
of saliency and/or sahuation.the generator acts as a frequency I=- m
converter as well as an impedance;in this case a more accurate
The linearised, general form of eqn. (6) can be written as
solution req- further modificationsand more detailed repre-
sentation [l].

During the iterative process of the Harmonic Domain solu-


tion the intemal emf's, derived from the initial load flow, re- Where
main constaut but these can be adjusted by further load flow
iterations as explained later in the paper.

Finally, considering the relatively small harmonic content


of the voltage waveforms, it is posible to linearize the behavi -
our of the non-linear annponents in the vicinity of the base
operating point. This process is discussed next.

ANALYSIS IN THE HARMO N C DOMAIN

The Linearisation Process

The non-linear function y = f(x) is transformed to the pha-


y=[F] X + YN
Y Y

sor form by a linearisation process in the m=


Harmonic Domain.The basic procedure is as follows
Consider the non-linearrelationship
y = f(x) (3)
where x and y are periodic variables, which when determined
by infinite Fourier series are expressed as follows Iviatrix of eqn. (8) is the Jacobian associated with eqn.
(5). It is, in structure, a Toeplitz matrix having the harmonics
in complex-conjugate pairs and the coefficients being of the
*
form C-i = Ci ,as shown by eqn. (9c).

Let us assume that the linearisation process takes place about


an operation point X b Y b i.e.

Let X and Y be the vectors of Xh and Y k Thus, eqn. (3) takes


the form
Y = f(X) (5)
Substitution of eqns. (10) into eqn. (8)gives
If eqn. (3) is differentiable, then for small increments about
base values XW yb the following relationship is valid
--
y=[F] x + YN (11)

Ay = f '(Xb) A X ( 6)

The terms of eqn. (6)are defined as -


YN is a vector of Norton harmonic injections.

Equation (11) can be interprekd as a Norton harmonic qui -


valent and represents the linearisation of eqn. (3) in the
Harmonic Domain.

In principle every non-linear component should be amena-


ble to a Norton eauivalent remesentation in the Harmonic
435

Domain. The main non-linear components of the conventio- the linearisation process described in the previous section. The
nal power transmission system have already been modelled in following change of variables is quired if the variables of the
the new Domain ;these include the salient-pole spchronous general equation (8)are used to represent the entire power net-
generator [l]. the single core [2] and multi-limb [3] power work.
transformers. Harmonic Domain models have also been den - G--A T
ved for the Static Var Compensator [4] and the three-phase
bridge rectifier [SI.

Similar techniques can be used to develop harmonic do- z+-


AT
main models for other non-linear components, such as large
induction motors. Equation (8)has now the general form

These Norton harmonic equivalents are combined with the E'=[Y,]zY


linear part of the network in a unified algorithm described in W h
the following section.

Structure of the HD Algorithm


- : harmonic admittance matrix for the entire network.
LyJ1
AV : vector of incremental voltages.
N

AI : vector of incremental currents.


The nodal r e p e n t a t i o n of the linear components as trans -
fer admittances is well established [6]. For use in the
The nodal admittance matrix of eqn. (13) contains all the
Harmonic Domain these admittances are represented in com-
network b u s h , phases and the selected spectrum of harmo-
plex-conjugate pairs, i.e. for a particular harmonic order i h
nica (positive and negative).
the following relations apply to the passive nodal mmpo-
nents:
* The basic nodal analysis used to form single and three-
Yself (- h) = Yself (h) phase admittance matrices is applied to determine the admit-
* tance matrix of eqn. (13). The order of each building unit of
ymutual (- h) = Ymutual (h) the non-singular matrix PJ]is equal to the number ofpha-
The linear network and linearized harmonic Norton equiva- ses * (2 * number of harmonics + U% rem) and is, in addition,
lent of the non-linear components are solved together by the structurally sparse.
Newton-Raphson procedure as illustrated in the block diagram
of Figure 1. A simple circuit, shown in Figure 2, is used to illustrate
the structure of the admitance matrix in the Harmonic
I INFlJC BASEVOLTAGE R O M
IDAD,WJW -1 Domain. It consists of 3 nodes and 3 phases and, to reduce the
display of information, the admittance matrix only shows the
fundamental and 3rd harmonic, as well as a d.c. component.
OBTAIN CURRFNT
HARMONICS AND M E l R
DWVATNB The diagonal blocks corresponding to the nodes with non-
I Ilinearities (nodes 2 and 3 of Figure 2) show in dots the ele-
FORM HARMONIC NORTON
ments of their Jacobian matrices. The remaining dots indicate
EQUIVALENTSANDCOMBINE
THFMWWHTHEUNUR
the presence of an admittance.
NEIWORKCOMPONB4TS
IThe linear and linearised components of the power network
SOLVEFORHARMONIC are assembled together to obtain a unified solution of the
VOLTAGES USING THE
NEWTON-RAFWSON MmHOD complete system. An iterative solution is, therefore, of neces-
sity. The linearisation, at each iterative step, of the non-linear
components implies the relinearisation of the complete sys-
tem through a Newton-type procedure, where Ty J] plays the
role of a Jacobian.

A full Newton-Raphson solution is not obtained because


part of the system is already linear and only a limited number
of harmonica is represented. On the other hand, voltage and
current excitations are applied to the network so that a partial
Figure 1. Harmonic Domain Algorithm inversion of Ty J] is required. A hybrid algorithm has been de-

The harmonic representation for the complete power sys- veloped for this purpose [7] that exploits efficiently the inver-
tem can be achieved following the rules of nodal analysis and se of the matrix FJ].
436
2 3
I I
I I

1 3
A B C

(b)
Figure 2. System representation in the Harmomc Domain
(a) Circuit diagram. (b) Structure of the Jacobian-admittance matnx

The iterative solution of eqn. (13). started from the base the repetitive evaluation of numerically sensitive Fourier coef-
case given by the threephase load flow solution, involves the ficients, roundoff errors may result from the multi-evaluation
reevaluation and partial re-factorisation of Ty J] at each itera- of [yJ] in the presence of considerable harmonic distortion in
tion. The computation time required is proportional to the the power network. The phase-angle and magnitude of harmo-
spectrum of harmonics represented, the size of the system and nic voltages and currents would be affected before the solution
the number of iterations needed to obtain the solution; the last is obtained for the complete network.
aspect depending on the distortion effect, produced mainly by
non-linear components. From cornprebensive sensitivity studies, carried out in refe-
rence [14] with a convergenceCriterion of O.ooO1 p.u. toleran-
It has been observed that the multi-evaluation of Ty Jl leads ce, the number of iterations varied between 7 and 10. Raising
to identical solutions as in the case when this mavix is only the hannonic content by increasing transformer saturation le-
evaluated at the first iteration. In fact with the last approach, vels only resulted in one or two further iterations. So far, the
besides efficient computation, faster convergence is obtained technique proposed has proved to be robust and has shown no
in most cases, moreover, it should be noted that considering convergenceproblems, even in cases of pronounced distortion
the size of the problem to be solved for a practical system and [31[131[141.
431

Case Study this effect is particurlarly noticeable at the Tiwai bus (Figure
4 f) where the 5th harmonic level in one phase exceeds the re-
The test system, shown in Figure 3, is a small part of the commended level by a factor of 7.
New Zealand South Island 220 kV network which contains a
static converter at Tiwai.
-
/'
Field tests carried out at Tiwai [SI had already shown the c:
presence of a parallel resonance at a frequency lying somewhe-
re between the fourth and fifth harmonic. This observation is
not exclusive to Tiwai but to a greater or lesser extent to se- - i30, ' '

0
-8
-hG e
veral of the network busbars.
i L

To investigate this problem and assess the background har- 37


monic noise in this region, the large converter plant at Tiwai
has been removed from the model. The generator and load I' , aF3
non-linearities have also been ignored, the only harmonic 5 -a 7[... .:

sources being the magnetizing currents of the five transfor- ,:


c
-h c
mers.
: ?[:
-
Pir
The loads are as specified in the diagram and the generators
internal e m f ' s are set at 1 PA., with the Mauapouri power sta-
tion acting as the slack bus. Details of the liuear and non-line-
ar components of the test system can be found in reference 9 -365: :
I I '
r31. ' 5

-
/'

9 L 8 ;i 16 2 j ' - 2 j i 5 6 5
time, r n S hCrrnOiiC c-;?
ii h

Figure 4.Voltage waveforms and related harmonic content


of test system with 1 p.u. internal emf's.
a, e Manapouri. b, f Tiwai. c, g Invercargill. d, h Roxburgh.
- Phase A. - - - Phase B. .... Phase C.

While the load flow condition selected may be considered


unrealistic, the studies show the importance of representing
the conventional system non-linearities; without these the test
system results would have been perfectly sinusoidal.

RFDUCFD W O R K REPRESENTATION

To alleviate the memory requirements of the Harmonic


Figure 3. Test System Domain it is advisable to reduce the size of the problem, retai -
ning only the non-linear nodes. This is justified on the basis
Figures qa-d) illustrate a full period of the voltage wave- that the doesn. need explicit representation
forms obtained at relevant network busbars and figures q e h ) of the part of the network, which is passive at harmonic
their harmonic content' the busbars show the frequencies. A reformulation of the equations is required, as
presence of considerablefifth harmonic. follows:
The IEEE recommended level for the individual harmonic
voltages is 1 %. It is clear from Figures qe-h) that this level
is exceeded by the 3rd and 5th harmonics at all the busbars;
438

&
(*) => Non-linear components
(**) 3 Iinearcomponentes

-vz
solving for V2 in equation (14)

= [Y22]-1 (iz - [YZl] Vi) (15)


INFWI DATA

f1= ([Yli] - [Yn] [Y22]'1 [YZiI) Vl + [Y 14 [Y22].1

(16)
i2
,
Finally, solving for V 1 in equation (16)
-
VI= [zl] - KO h) (17)
UNFARIZE EACH NON-UNFAR
Where COMFUNENT W "HE HARMONIC
WMAW AND FORM ITS NORmN
E Q I J I V A W CIRCUIT

(
[z11= [Yll] - [Y 121 [y22]-1 [YZl] ) - I (1%)

KO = [Y 123 [Y22]-i
Once the non-linear node voltages are obtained from eqna-
(la)
I I COMBINEUNEAR AND UNFARlZED
NFTWORKS, SOLVETHE NU
LNFARIZEDSYSTEMTO
CAUlJLATEHARMONIC VOLTAGES
1
tion (17). their substitutionin equation (15)permits the cal-
culation of the linear node voltages. In the absence of any cu-
rrent injection in the latter nodes (i.e. I2 = 0) the solution of
equation (15) is very simple.

Finally, the matxix product m21] only needs to be


calculated once for use in the equations (15)and (18).

LOAD FLOW AND HARh4oNIc ANALYSIS


figure 5.Unified Harmonic Domain Solution

&
The Harmonic Domain algorithm provides an accurate so-
lution of the interaction that exists between the fundamental
and harmonic frequencies. In the process, however, the soluti-
on following convergence requires fundamental voltages
READ DATA
slightly different from those specified by the load flow. If the
originally specifiedgenerator voltages are to be maintained, a
new load flow and harmonic solution needs to the carried out.

An algorithm to solve the load flow and harmonic flows


simultaneonsly by the Newton method was described by Xia
and Heydt [9]for the case of a static converter connected to a
symmehical power system.

Such algorithm has recently been extended to the three-


phase load flow and applied to derive the harmonic content
produced by a Static Var Compensator [IO] [ll].The structure
of the simultaneous algorithm is shown in the flow diagram
of Figure 5. However the generalization of the simultaneous
three-phase load and harmonic flow for use in a system with (&)
n (Ti&-)
I
I
multiple non-linearities is a problem of great complexity.

Considering the different requirements of the load flow and


harmonic studies, it seems more appropriate to solve the two
parts sequentially, taking advantage of the simplified harmonic
solution described in the previous section.The structure of the w
resulting sequential algorithm is shown in Figure 6. Figure 6. Sequential Harmonic Domain Solution
43 9

It involves a double iterative scheme, which starts with the CONCLUSIONS


three-phase load flow (TPLQ. Upon convergence of the load
flow the nodal information is converted into internal emf's The structure of the Harmonic Domain, a new frame of re-
and equivalent nodal admittances, as explained earlier. This is ference for the harmonic analysis of power systems has been
followed by the Harmonic Domain (HD) solution, at the end described. In this domain the network busbars and coupling
of which the load flow and harmonic balance tolerances are between phases aad between harmonics are explicitly represen-
checked to decide whether another double iterative cycle is re- ted.
quired
In this frame of reference the harmonic sources are repre-
This effect has been quantified in a recent paper [ 121 with sented as Norton equivalents whose components vary during
reference to the harmonics caused by an aluminium smelter the solution. However, the limited experience gained so far in-
connected at the Tiwai bus of the test system of Figure 3. The dicates that the Jacobian elements of the Norton admittance
smelter is represented as a 6-pulse convertor load rated at 75 matrix vary little from iteration to iteration and it is rarely ne-
Mw, plus a F'Q load of 250 MW and 60 MVAr. This consti- cesary to recalculate them.
tutes a worst-case situation when one rectifier unit is removed
for servicing and the presence of some voltage imbalance cau- For the calculation of the non-characteristic harmonic in-
ses uneven load sharing among the remaining units. jections it is advisable to use the double iterative scheme, i.e.
adjust the load flow following the Harmonic Domain itera-
In this cases filters are provided for the 5th. 7th. 11th and tions.
13th harmonics, and supply approximately 22 MVAr at fun&
mental fquency. This general frame of reference has been applied to the har-
monic solution of a practical transmission system with deta-
The main results, repeated in Table 1 , show that the load led representation of the transformers' magnetising non-linea-
flow is practically unaffected by the presence of harmonics; rities. The use of the Harmonic Domain at the planning stage
this is because the characteristic harmonics are absorbed by should help to assess the best practical value of the transfor-
the filters. mer magnetising characteristic(s) and thus improve the overall
economy of a particular system extension.

Quantity Initial load flow Final load flow With the system already in operation, this algorithm will
PI 24.9585 24.9268 (- 0.13%) help to assess the influence on the network harmonic distor-
p2 25.4074 25.4465 (0.15%) tion of existing nodal voltages and current levels, so that a re-
p3 25.6074 25.6224 (0.06%) distribution of harmonic flows by alternative control actions
Ql 12.5151 12.5132(- 0.02%) in real-time could be attempted when required.
B 12.7772 12.7579(- 0.15%)
4 12.4450 12.4753(0.24%) Further work is needed to develop Harmonic Domain mo-
d' 2.05896 2.05911 (0.007%0) dels of large power electronic components, such as HVDC
Id 1.10698 1.10698 converters.The establishment of accurate initial settings parti -
d' 2. 27923 2.27939 (0.007%) cularly for the non-characteristicharmonics, is another area in
need of attention.
Table 1: Comparison of load flow results
REFERENCES

[l] Semlyen, A; Eggleston, J F & Arrillaga, J (1987):


Four load flows Admittance matrix model of a synchronous machine for har-
monic analysis. IEEE Trans. on Power Systems, Vol PWRS
0.000561(29%) 2(4), pp. 833-840.
0.000399 0.000509(28%)
0.000337 0.000417(24%) [2] Semlyen, A; Acha, E & Amllaga. J (1987): Harmonic
0.000260 0.000306(18%) Norton equivalent for the magnetising branch of a transformer.
Proc. IEE Pt. C, Vol 134(2),pp. 162-169.
Table 2 Comparison of triplen harmonic current
injections [3] Acha, E; Arrillaga, J; Medina, A & Semlyen, A
(1989): General frame of reference for analysis of harmonic
On the other hand the non-characteristic harmonics can be distortion in system with multiple transformer non-linearities.
greatly affected by the adjusted load flows, as shown in Table Proc. IEE Pt. C, Vol 136(5),pp. 271-278.
2, which contains the triplen harmonic content in one phase
after the first and final cycle of load flow and Harmonic 141 Xu, W and Donunel, H W: Computation of steady
Domain iterations. state harmonics of Static Var Compensator. ICHPS 111,
440
Nashville; In, 1988. University of Toronto, Canada.

[5] Valcarel. M A d i s i s del rkgimen v e n t e de 10s Maria L. V. Lisbon received an engineering degree in
sistemas el6ctricos de potencia con elementos no lineales me- Brazil in 1982, where she practiced as an ElectricalEngineer
diante un mktodo & reparto de cargas con arm6nicos. Tesis with Eletrobrh until 1992 when she joined the University of
Doctoral, Universidad Politecnicade Madrid. 1991. Canterbury (New Zealand) as a Ph. D. student.

[6] Arrillaga. J; Bradley, D & Bodger. P S: Power System Maria A. Cavia was born in Santander (Spain), received
Harmonics. John Wiley & sons. London. 1985. her degree in physical electronics from the University of
Cantabria (Spain) in 1980, where she has been a lecturer since.
[7] Medina, A, Amillaga, J & Acha, E (1990): Sparsity- 1986.
oriented hybrid formulation of linear multiports and its appli-
cation to harmonic analysis. IEEE Trans. on Power Delivery, Paulino SPnchez received his degree in physical electronics
V015(3) pp. 1453-1458. from the University of Cantabria (Spain) in 1976. He obtai-
ned a PhD in Electrical Engineering from the University of
[8] Densem, T J; Bodger, P S & Arrillaga, J (1984): Three Cantabria, where he is a professor.
phase transmission system modelling for harmonic penetra-
tion studies. IEEE Trans. on Power Apparatus and Systems,
Vol PAS 103(2), p ~310-317.
.

[9] Xia, D & Heydt, G T (1982): Harmonic Power Flow


Studies. Part. 1. Formulation and Solution. IEEE Trans.
PAS-101.6, pp. 1257-1265.

[lo] Xu, W; Marti. J R & Dommel, H W (1991): A mul-


tiphase harmonic load flow solution technique. IEEE Trans.
PWRS Vol6, No 1, pp. 174-182.

[ l l ] Xu, W; Marti, J R & Dommel, H W (1991):


Harmonic analysis of systems with static compensator. IEEE
Trans. PWRS Vol6. No 1. pp. 183-190.

[12] Arrillaga. J; Callaghan, C D (1989): Double-iterative


algorithm for the analysis of power and harmonic flows at
a.c.-d.c. convertor terminals. Roc. IEE Vol 136, No 6. pp.
319-324.

[13] Medina, A; Arrillaga, J (1993): Harmonic interaction


between generation and transmission systems. IFEE Trans.
PWRD, pa~er93 W M 059-6.

[14] Medina, A (1992): Power Systems Modelling in the


Harmonic Domain. Ph.D.Thesis, University of Canterbury.
New Zealand

AUTHORS

Jos Arrillaga received his BE degree from Spain and his


PhD and DSc from Manchester, where he led the power sys-
tems group of UMIST between 1970-74. He has been a pro-
fessor at the University of Canterbury since 1975. He is a
Fellow of IEE, IEEFi, the New Zealand Institution of
Enginem and the Royal Society of New Zealand.

Aurelio Medina received his BE from Mexico and his PhD


from New Zealand. He worked for the Mexican national uti-
lity and the Instituto de Investigaciones El&ricas, Mexico,
respectively. At present he is a Post-Doctoral Fellow at the

Вам также может понравиться