Sei sulla pagina 1di 4

NEWTON-RAPHSON PROBABILISTIC HARMONIC POWER FLOW

THROUGH MONTECARLO SIMULATION


S. Nos M., Member IEEE R. Castafieda P.
Department of Electrical Engineering, Catholic University of Chile
Vicuiia Mackena 4860, Santiago, Chile
e-mail: sriosaing puc.cl, ramon@pfloyd.ing .puc.cl
I

Fax: (56) -2 - 552 2563, Phone: (56) - 2 - 6864284

Abstruct-A method to assess probability distributions of A method to assess probability distributions of nodal
nodal harmonic voltages and currents in electrical networks is harmonic voltages and currents is presented in this paper.
presented. The method considers a deterministic model for the The method considers a deterministic model for the
converter equations, that are solved through a Newton- converter equations, that are solved through a Newton-
Raphson hamonic power flow. Based on Montecarlo simulation Raphson hamonic power flow. Based on Montecarlo
the randomness of both the electrical network variables and
converter operation parameters is incorporated in the simulation, the randomness of both the electrical network
calculation of probabilistic harmonic voltages and currents. A variables and converter operation parameters is incorporated
simulation example that considers a harmonic filter installation in the calculation of probabilistic harmonic voltages and
in an industrial plant, is included. Also, deterministic and currents. A simulation example that shows the main features
probabilistic approaches are compared. of the stochastic approach and its advantages over
deterministic approach is included.
I. INTRODUCTION
11. PROBLEM
STATEMENT
The assessment of harmonic levels in transmission,
distribution networks, and industrial plants, is nowadays a The application of probabilistic methods to assess
very important issue for both, utilities and customers [l], [2]. harmonic levels in electrical networks have been considered
Deterministic approaches based on: time domain analysis by several authors. By resorting to analytical methods [SI,
[3]; frequency domain linear methods [4], [ 5 ] ; frequmcy [9], numerical integration [lo], convolution [ l 11,
domain non-linear analysis, considering both balanced [6] Montecarlo simulation [lo], and semiempirical methods
and unbalanced networks [7], have been published. [12], [13], the probability density functions of the network
Since a long time ago, the nature of harmonic problems harmonic voltages are calculated.
has been recognized to be essentially stochastic, therefore an Montecarlo simulation methodolcigy has shown to be one
accurate assessment of harmonic levels in an electrical of the most effective [lo] to assess the stochastic
network will require the application of probabilistic characteristics of harmonic voltages and currents in
methods. electrical networks.
Some specific applications where a probabilistic In [lo] simplified models of non-linears loads are used.
assessment of harmonic levels is required: Harmonic propagation through the network is calculated by
a frequency domain linear method.
Harmonic filter design In this work a detailed representation of the static
Harmonic standards and recommendations converter is used. It takes into account harmonic component
Calculation of maximum voltage applied to capacitors interaction through a Newton-Raphson harmonic power
Expected power losses in transmission, distribution lines flow. In addition, probabilistic models of main converter
Telephone interference parameters are implemented using Montecarlo simulation.
Others
111. APPLICATIQN OF MONTECARLQ SIMULATION
METHODQLQGY

Probability distribution functions were considered for the


following variables:

Converter active power (Pc), effective current (Ic), and


firing angle (a,)
Unbalance (percentage of negative sequence voltage) in
0-7803-2972-4/96$5.00@1996IEEE equivalent voltage supply (U)

3 297
By defining E the vector whose componentes are the above
variables (Pc, IC , ac, U). Vector E components have a
(rkl=f(vI, v2, * * f vhl (1)
For the k-th harmonic voltage:
stochastic behavior that is reflected in the harmonic voltages
and currents of the electrical network.
(Vk) -(Z". Ik
=(T)YI:J (2)
Fig. 1. shows the flow diagram of Montecarlo simulation. Where [Zk] is the node impedance matrix. The first term
A particular value of E is obtained from the probability on the right hand side represents the "open circuit" voltage.
distributions associated to each E - component. Next, this Vector Ik represents harmonic currents injected by
value is used to calculate harmonic voltages and currents in converters.
a non-linear (Newton-Raphson) frequency domain
deterministic program. Resulting harmonic currents and
voltages are stored, and a new simulation is performed until
the number of Montecarlo simulations (5000, in order to
obtain statistically valid results) are completed.

2iiL-<I
START

OPERATING CONDITIONS
S,MU~TION PROBABILISTICMODELS

Fig. 3. Voltage and current waveforms

I HARMONIC PROPAGATION
SIMULATION
-<I DETERMINISTIC MODELS
A Newton-Raphson algorithm is implemented in a
computer program for the calculation of harmonic currents
and voltages. Using the admittance matrix, the harmonic
voltages are calculated for the rest of the electrical network
nodes.
Both balanced and unbalanced networks (modeled
through symmetrical components) are considered. The
following unbalances have been implemented:

Unbalances in converter transformer impedances


~

Unbalances in firing angles of converters


Fig. 1. Flow diagram of Montecarlo simulation Negative voltage sequence in equivalent electrical
network
IV.CONVERTER
AND NETWORK
MODELLING
V. SIMULATION
EXAMPLE
A non-linear frequency domain model is used [ 6 ] .171 that
consider an equivalent system as in Fig. 2. An 11-bus system, with two six-pulse 7.5 MW (with a
system power base of 10 MVA) converters in buses 10 and
I I 11, is shown in Fig. 4. Capacitor banks can cause harmonic
resonances, thus a filter installation is considered. Filter
performance will be evaluated through the average voltage
THD (AV.THDv) of the whole network buses.

Deterministic approach
Fig. 2. Converter and Thevenin equivalent network models A deterministic harmonic power flow without including a
filter was performed and the Av.THDv obtained is 6,559%.
Electrical network is reduced to a generalized Thevenin
It can be seen in the first column of Table 1 that buses 8, 9
equivalent using the compensation method. Converter
and 11 present the greater THD indexes (especially due to
equations are solved considering six conduction periods as
5th and 7th harmonics values).
shown in Fig. 3.
A passive single tuned filter is then installed, with
Harmonic current of order "k" injected by cmverter
impedance lor the "k-th" harmonic given by equation 3:
depends upon all harmonic voltages:

1298
k 1 l k l (3) 0.675 and 0.825 p.u. Table 2 summarizes the results
(---->
Zf = Bc'Ko'QcaI+j.-.
BC K~2 k obtained. The Av.THDv (X)for three deterministic cases
are shown in the first three columns. In the last column the
"expected values" from the probabilistic analysis are shown.

Table 1. Individual and average THD for different filter locations

Total Harmonic Distortion


3.8kV

%i
Filter :ation

mt
Bus Without Bus 9 Bus 10 Bus 1 1
filter
1.05 MVAr 2x1.8 MVAr 0.75MVAr
1 3.70 3.49 3.51 3.49
3.6 MVAr
2 3.77 3.55 3.58 3.55
3 5.39 5.36 4.90 5.36
4 4.04 3.83 3.85 3.83
5 5.85 5.34 5.63 5.34
6 4.40 4.21 4.21 4.21
4.16 kV

G
Fig. 4. 1 1-bus system with two converters
7
8
9
10
11
Average I
6.67
9.92
1 1 .09
6.21
11.11
6.559 I
6.12
8.84
9.83
6.24
9.86
6.061 I
6.43
9.65
10.80
5.61
10.82
6.272 I
6.12
8.84
9.93
6.24
9.84
6.059

Where:
Bc : Inverse of Qfc = MVAr of filter capacitor
KO : Resonance frequency Table 2. Individual and average THD variation with active power P
Qcal : Quality factor of filter
Deterministic Analysis Stochastic
Different filter locations within the network were Analysis
examined. The following parameters for the filter: Qfc=2 Bus P = 0.675 P = 0.75 P = C.825 P = U [0.625,
MVAr, Qcal=5, Ko=550 Hz, were considered. 0.8251
Table 1 shows the results obtained. The better Av.THDv 1 3.04 3.29 3.52 3.31
(6.059) occurs with the filter connected to bus 11 where one 2 3.09 3.35 3.57 3.36
converter is located. 3 3.83 4.12 4.38 4.14
It is interesting to observe that the effect of locating the 4 4.30 4.65 4.96 4.67
filter in bus 9 (Av.THDv=6.061) is greater than locating it 5 4.54 4.93 5.29 4.96
in bus 10 (Av.THDv=6.272), although a converter exists in 6 4.01 4.31 4.57 4.33
bus 10. When the filter is located in bus 9 or 11 it "isolates" 7 5.44 5.87 6.25 5.89
the two converters, thus reducing the interaction between 8 4.87 5.27 5.64 5.29
harmonic currents, and attenuate harmonic injection of the 9 11.47 12.51 13.47 12.58
4.16-kV converter. 10 11.63 12.71 13.72 12.78
On the other hand, when filter is located in bus 10 (see 11 11.64 12.73 13.73 12.79
Table I), the 4.16-kV converter harmonic injection Average 6.17 6.70 7.19 6.74
propagates through lines between buses 9-8 causing higher
THD levels in all buses except buses 3 and 10.
It can be observed that the deterministic simulation in two
Stochastic Analysis cases under estimates the Av.THDv This can lead to a
mistake in filter design causing an overload to the filter.
For the network in Fig. 4, but considering only one Fig. 5 shows the probability functions for fwndamental
converter connected to bus 11, a probabilistic harmonic harmonic voltage at the bus 11.
power flow is performed considering the active power ds a
stochastic variable. The density function is uniform between

1299
400
VIII. REFERENCES
350
300 [I1 IEEE Power System Harmonic Working Group, "Bibliography
250 of Power System Harmonics Part I and II", Papers 84WM
Number of 200 214-3, IEEE PES WinterMeeting, Jan. 29-Feb. 3, 1984.
ocurrences
150 PI J. Amllaga, D. A. Bradley y P.S. Bodger, Power System
Harmonics, John Wiley & Sons, July 1989.
100
[31 H. W. Dommel, "Digital Computer Solution of

'
50 Electromagnetic Transients in Simple and Multiphase
0 Networks", IEEE Transactions on Power Apparatus and
0 94669 0 95028 0 95386 Systems, Vol PAS-88, No, 4, April, 1969, pp. 388-389.
[41 A. A. Mahmoud, R.D. Shultz, "A Method for Analyzing
Fig. 5 . Probability distribution of fundamental harmonic voltage Harmonic Distribution in A.C. Power Systems", IEEE
Transactions on Power Apparutus and Systems, Vol. PAS
101, No. 6, June 1982, pp. 1815-1824.
It can be observed that this voltage "follows" the uniform [51 J.P. Tamby, V.I. John, "Q-HARM - A Harmonic Power Flow
distribution of active power. In Fig. 6 the same graphic is Program for Small Power Systems", IEEE Transactions on
depicted for the 1lth harmonic voltage. Power Systems, Vol3, No 3 , August 1988, pp. 949-955.
D. Xia, G.T. Heydt, "Harmonic Power Flow Studies. Part I -
Formulation and Solution", IEEE Transactions on Power
1400 Apparutus and Systems, Vol. PAS 101, No. 6, June 1982, pp.
1257-1265.
M Valcarcel, J.G. Mayordomo, "Harmonic Power Flow for
1000 Unbalanced Systems", IEEE Transactions on Power Delivery,
06 Vol 8, No 4, October 1993, pp. 2052-2059.
Number of Probability
ocurrences 600 distribution S.R. Kaprielian et al., "Predicting Voltage Distortion in a
04 System with Multiple Random Harmonic Sources", IEEE/PES
400 03
1994 WinterMeeting, New York, USA January 30 - February
02
200 3, 1994.
01
W.E. Kazibwe, T.H. Ortmeyer, M.S.A.A., "Summation of
0 0
0 013 0 013 0 013 0 014 Probabilistic Harmonic Vectors", IEEE Transactions on
Power Delivery, Vol4, No 4, January 1989, pp. 621-628.
Y. Baghzouz, 0. T. Tan, "Probabilistic Modeling of Power
Fig. 6 . Probability functions for 1lth harmonic voltage System Harmonics", IEEE Transactions on Industry
Applications, Vol IA-23, No. 1, Jan.-Feb. 1987, pp. 173-180.
Obviously the distribution is not uniform. In fact the [I11 M.R.G. AI-Shakarchi, K.A.R. AI-Anbarri, "Power Systems
Probabilistic Harmonic Load Flow", Proceedings of Fourth
mayor ocurrences are in the greater voltages range, thus
International Conference on Harmonics in Power Systems,
explaining results given in Table 2, i.e., the harmonic Budapest 1990, pp. 63-69.
voltages have expected values for 5000 simulations greater [121 J.M Crucq, A. Robert, "Statistical approach for harmonic
than the values of a simple deterministic simulation. measurement and calculation", CIRED, Rep. 2-02, 1989, pp.
91-96.
VI. CONCLUSIONS [I31 A. Cavallini, M. Cacciari, M. Loggini, G.C. Montanari,
"Evaluation of harmonic levels in electrical networks by
statistical indexes", Con$ Record of the IEEE IAS 28th
An algorithm to assess probability distributions of nodal Annual Meeting, Ontario, Canada, Oct. 218 1993, pp. 1586-
harmonic voltages and currents in electrical networks, and 1593.
an application to a 11-bus network, have been presented.
The method allows to calculate stochastic harmonic Sebastian Rios is a Member of IEEE. Currently Associate Professor
at the Faculty of Engineering, Electrical and Electronics Engineering
currents injected by pulse converters into the electrical
Departament at the Catholic University of Chile. Prof. Rios obtained his
network. M. Sc. in Electrical Power Engineering at Heriot-Watt University, and the
In the example application the results show that a Ph. D. degree from the University of Manchester Institute of Science and
deterministic analysis can under estimates harmonic voltages Technology (UMIST), Great Britain. His research includes the
caused by converters. Also it is shown the strong interaction assessment of the Quality of Electrical Service for G, T & D utilities,
Dynamic Simulation of Power Systems, Reliability and Security Analysis
between harmonic components injected by converters, which of electrical networks.
emphasizes the importance and use of detailed converter Ramon Castaiieda obtained his M. Sc. in Electrical Power
models. Engineering from Catholic University of Chile. He has been Research
Assistant at the Electrical and Electronics Engineering Departament, and
VII. ACKNOWLEDGEMENTS has been involved in research projects in the area of Quality of Electrical
The authors wish to thank Catholic University of Chiie and Service. His actual research is the application of probabilistic
FONDECYT grant 1950179. methodologies to harmonic analysis.

1300

Potrebbero piacerti anche