Vous êtes sur la page 1sur 6

PREDICTION OF HARMONIC DISTORTION DUE TO THE APPLICATION OF CAPACITORS TO UNBALANCED THREE-PHASE SYSTEMS.

M.J.Bridgeman, ALSTOM: Protection and Control, United Kingdom R.E.Morrison Monash University, Australia Norman MacLeod, ALSTOM: Power Electronic Systems, United Kingdom S.B.Tennakoon, Staffordshire University, United Kingdom

Abstract: This paper considers the application of plain capacitors to a system with unbalanced harmonic voltage distortion. The magnification of existing harmonic voltages may result following the connection of the capacitors. However to predict the resulting voltage distortion it is necessary to consider the system damping due to any converters on the system. This paper covers the modelling of damping due to shunt loads including the converter. It is demonstrated that frequency domain models are capable of predicting the resulting voltage magnitudes following the connection of a capacitor providing a realistic model is used for the converter and shunt load. Keywords : Harmonic distortion, Reactive power
control, Resonance.

1. Introduction
Harmonic voltage distortion is often due to the flow of non-sinusoidal current through system impedance. Most systems contain some distorting loads with the result that few network voltages are distortion free. However when voltage distortion reaches particular trigger levels interference and damage may result to some sensitive loads. To minimise potential problems voltage distortion is normally maintained below limiting values defined by appropriate standards. For a given distorting load higher distortion voltages result from high system impedance. Many power systems contain capacitance which is present for power factor correction. Unfortunately the interaction between parallel capacitance and series inductance leads to a parallel resonance producing a relatively high impedance at a small range of harmonic components. The net result of this enhanced impedance is excessive voltage at a few harmonic components having frequencies close to the resonant frequency. When resonance occurs the magnification of voltage distortion depends on the sharpness of the resonance as determined by the system Q factor or X/R ratio. The X/R ratio is strongly influenced by the circuit resistance contained in series form in the supply network and shunt form in the loads. Although the resistance is difficult to estimate the value is important when assessing the likelihood of potential problems. Knowledge of the X/R ratio becomes critically important when applying system capacitors. Although the option exists to place so called detuning reactors in series with the capacitors to

avoid resonance this option is expensive. However calculations to evaluate the effect of damping are affected by the accuracy of the circuit resistance. Because the correct level of damping is often unknown, calculations to determine whether detuning reactors are needed often predict a level of distortion voltage which is never realised in practice. Some environments are known to have significant damping effects on parallel resonance such that detuning reactors are not normally needed for power factor correction units. An example of this environment is a large building which may have a switched centrally connected power factor correction capacitor. Measurements of harmonic distortion within buildings often confirm that little voltage amplification has taken place due to the capacitor. It is probable that much of the damping effect has been generated by system loads which are in practice large numbers of relatively small loads. Unfortunately little is known about the damping due to large numbers of small loads and models to represent the damping are not well developed. This paper considers damping due to loads including rectifier load. The effect of unbalanced rectifier load on voltage enhancement is considered by the use of symmetrical components applied at a range of harmonic frequencies. A time domain simulation of a simple Power System exhibiting significant levels of harmonic current imbalance is used to demonstrate the presence of damping due to rectifier and other load. The load impedance is evaluated using an algorithm originally applied to measure system impedance during site tests. The extension of the algorithm to include the calculation of the impedance

including the resistance at non-characteristic harmonic frequencies is described. Fig. 1 shows a single line diagram for the simple unbalanced system under simulation. A cable model was included at 11kV in order to introduce a parallel resonant characteristic into the system harmonic impedance. The system contains a switched capacitor bank used for power factor control. Three singlephase current controlled converters produce unbalanced harmonic current distortion. The converters were set to produce DC currents of 200, 250, and 300 Amps respectively. A suitable singlephase converter model is not available in EMTDC and thus the firing circuit for the fully controlled bridge rectifier had to be implemented using standard library components [2]. The capacitor and system impedance produced a potential parallel resonance at the 7th harmonic with all of the capacitors connected. The linear component of the system loading was chosen to offer moderate damping (equivalent resistance approximately 0.3 at 240V).

harmonic impedance utilises system voltage and current perturbations induced by capacitor switching, where

Zn =
The values

Vn I n
(1) are the changes in

Vn and I n

harmonic voltage and current sequence components resulting from a capacitor step change. The calculation of damping is achieved by considering a particular harmonic voltage before and after the capacitor switching operation. It may be shown that the circuit resistance may be derived from the following equations [1]. aKd 2 + bKd + c = 0 (2)
Kd is the reciprocal of the damping resistance, and

a = ( Rs2 + Xls2 ) ( Kv2 1)


(3) b = 2 Rs Kv 2 1

)
(4)

3.85mH 11 kV

10 kms of 300mm Single Core Cable

((X c=

c2

Xls ) + Rs 2 Kv 2
2

X c2

((X

c1

xl s ) + Rs2
2

Xcl

11 kV

(5)
Kv = harmonic voltage magnification, Xc2 = reactance of PF capacitor after switching. Xc1 = reactance of PF capacitor before switching. Rs = Resitive component of system impedance. Xls = Reactive component of system impedance.

1.25 MVA @ 4.5%

415 V

V
Iadc = 400 A Ibdc = 500 A Icdc = 600 A 100 kVAr 100 kVAr 100 kVAr 100 kVAr

600 kVA @ 0.8 p.f.

Although the method was originally intended for use with measurement systems it is applied here to a time domain simulation to derive effective resistance values for loads consisting of a converter in parallel with some linear or clean load. 3. HARMONIC IMPEDANCE EVALUATION The harmonic impedance calculation yields three impedance values for each harmonic, each one corresponding to a different sequence component. The 5th, 7th, 11th, 13th, 17th, and 19th harmonics were considered. In order to verify that the method was indeed producing the correct impedance the algorithm was first applied to the system impedance where

Fig. 1. Simple unbalanced system with switched capacitor unit. 2. DETERMINATION OF IMPEDANCE A simple method has been developed to extract system impedance on line from natural disturbances caused by the switching of a capacitor[1]. An extension of this method has been carried out to consider unbalanced currents by applying symmetrical components. The algorithm used for a balanced system was applied to each of the sequence components independently. The calculation of

comparison with alternative evaluation methods was possible. Simulating the injection of constant harmonic current into the system with the load removed derived the alternative impedance values. The voltage measured at the terminals of the current source indicates the system impedance at the particular harmonic frequency. This process was repeated for all of the harmonic sequence components under consideration. Fig. 2, Fig. 3 and Fig. 4 show, respectively, the positive, negative and zero sequence impedance loci derived from the current model. The positive and negative sequence impedance loci are identical since no rotating plant was present. The zero sequence impedance locus is a straight line based purely on the impedance of the supply transformer. The delta (HV) side of the star/delta transformer acts as a short circuit to zero sequence currents and thus this terminates the system impedance for the zero sequence equivalent circuit. No transformer resistance was included to enhance the effect of damping due to the loads.
0.35 0.3 0.25
17

0.18 0.16 0.14 0.12 19 17

X (Ohms)

0.1 0.08 0.06 0.04 0.02 0 -0.1 -0.05 0

13 11

7 5 3

0.05

0.1

0.15

0.2

0.25

R (Ohms)

Fig. 4. Zero sequence harmonic impedance locus. Fig.5, Fig. 6 and Fig. 7 show the impedance scatter plots at positive, negative and zero sequence, derived by applying the switching technique described in section 2. The different values of impedance at each harmonic resulted from the various capacitor switching perturbations. The scatter for all of the low order harmonic impedances, regardless of sequence component, was small, with a slight increase in scatter occurring for the 17th and 19th harmonics. It is believed that the increased scatter is due to the low magnitude of the higher order harmonics. The centre of the scatter for each of the harmonic impedances closely matched the simulated value of harmonic impedance, providing a good foundation upon which calculations could be performed to calculate damping due to the load. The presence of the scatter indicates that the method to determine impedance has imperfections. It is believed that the principal reason for the scatter is the difficulty in deriving a stable starting point to measure phase angles for harmonic voltage and current. The introduction of more capacitance raises the 415V system voltage and causes the real power absorbed by the clean load to increase. The flow of real power through the system transformer causes a small phase shift resulting in errors in the calculated value of the phase of harmonic voltages and currents. Clearly for a calculation based method the phase shifts could be accounted for. The effect is included here to indicate the difficulties in applying the technique to site tests. The problems would have been more severe, perhaps intolerable if the switching effect of a load carrying real power had been considered instead of a capacitor absorbing reactive power.

19

X (Ohms)

0.2 0.15 0.1


7 13 11

0.05 0 0

5 3

0.05

0.1

0.15

0.2

0.25 R (Ohms)

0.3

0.35

0.4

0.45

0.5

Fig. 2. Positive sequence harmonic impedance locus.

0.35 0.3 0.25

19

X (Ohms)

17 0.2 0.15 0.1 0.05 0 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 13 11 7 5 3

R (Ohms)

Fig. 3. Negative sequence harmonic impedance locus.

4. LOAD DAMPING RESISTANCE CALCULATION


0.35 0.3 0.25 17th Harmonic X (Ohms) 0.2 0.15 0.1 7th Harmonic 0.05 0 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 R (Ohms) 5th Harmonic 3rd Harmonic

19th Harmonic

13th Harmonic 11th Harmonic

The method was applied to the calculation of the real component of the load impedance [1]. Values of the sequence component of load resistance were calculated for all of the harmonics under consideration and are shown in tables 1, 2 and 3. At the higher order harmonics, the calculated values of damping resistance were close to the value expected from the active component of the system load (approximately 0.3 ). At harmonics 5 and 7 the load resistance was generally lower. While the exact reasons for the lower load resistance are not fully evaluated part of the reason may be additional damping effects due to the converter group. The calculation of resistance is an incremental calculation deriving affects around a particular operating point. For the particular operating point chosen the voltage current relationship at the terminals of the converter group may be resistive, even in a negative sense, depending on the conditions within the converter [3]. Therefore on some occasions the additional damping due to the converter may be helpful in reducing the effect of voltage magnification due to resonance.
Harmonic 5 7 11 13 17 19 Positive Sequence Load 0.1706 0.3067 0.3190 0.2644 0.3460 0.2606

Fig. 5. Positive sequence harmonic impedance scatter plot.


0.35 0.3 0.25 17th Harmonic X (Ohms) 0.2 0.15 0.1 7th Harmonic 0.05 0 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 R (Ohms) 5th Harmonic 3rd Harmonic

19th Harmonic

13th Harmonic 11th Harmonic

Fig. 6. Negative sequence harmonic impedance scatter plot.


0.18 0.16 0.14 0.12 X (Ohms) 0.1 0.08 0.06 0.04 0.02 0 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 R (Ohms) 13th Harmonic 11th Harmonic 19th Harmonic 17th Harmonic

Table 1. Calculated positive sequence damping.


Harmonic 5 7 11 13 17 19 Negative Sequence Load 0.2614 0.2814 0.2885 0.2958 0.2582 0.2613

Table 2. Calculated negative sequence damping.


7th Harmonic 5th Harmonic 3rd Harmonic

Fig. 7. Zero sequence harmonic impedance scatter plot.

Harmonic 5 7 11 13 17 19

Zero Sequence Load 0.1592 0.2332 0.2946 0.2936 0.2647 0.3323

Table 3. Calculated zero sequence damping.

5. UNBALANCED PHASE VOLTAGE PREDICTIONS Using the system without capacitance as a base condition a constant harmonic current model was derived for the converter group. The model was based on sequence components with each harmonic current group represented by positive, negative and zero phase sequence components. It was then assumed that the currents were constant, resulting from ideal sources, in a subsequent calculation of harmonic voltage distortion with the complete capacitor group connected. The resistances in tables 2, 3 and 4 were included in parallel with the ideal current sources creating more realistic sources. Having derived the sequence components of voltage distortion they were consequently recombined to form the phase values. A second calculation was carried out in the time domain to mirror the frequency domain calculation. The second time domain calculation was carried out in order to provide a reference for the frequency domain results. The seventh harmonic voltage from the time domain and frequency domain calculations is plotted in Figs 8, 9 and 10. The 7th harmonic voltage is considered since this one is most severely amplified by parallel resonance. It is clear from Figs 8, 9 and 10 that the simulated (frequency domain) results are close to the predicted (Time domain) results. The close agreement provides evidence that the imperfect frequency domain models used are a good representation of the converter and linear load. Table 4 presents the predicted and simulated voltages for all the harmonics considered. The values are sufficiently close to give confidence that the converter model is sufficient across a wide range of harmonics albeit for one set of operating conditions of the converter group. 6. CONCLUDING REMARKS A method has been proposed to derive an imperfect current model to account for inherent damping within converter systems. Calculations have been carried out to demonstrate the accuracy of the model when applied to a system to which a block of power factor correction capacitors is connected. By comparing the results from the imperfect frequency domain model to a time domain calculation it is concluded that the results are able to model the additional damping due to the converter. The results from this study indicate that additional damping may be resulting from the converter which

is acting to minimise the voltage amplification due to resonance. Further work will investigate the action of converters to control damping.
8.000 7.000 Harmonic Distortion (Volts) 6.000 5.000 4.000 3.000 2.000 1.000 0.000 Predicted Voltage 7th Harmonic Simulated Voltage 7th Harmonic Step 2 Step 3 Step 4

Fig. 8. Graphical representation of A phase 7th harmonic voltage.


8.000 7.000 Harmonic Distortion (Volts) 6.000 5.000 4.000 3.000 2.000 1.000 0.000 Predicted Voltage 7th Harmonic Simulated Voltage 7th Harmonic Step 2 Step 3 Step 4

Fig. 9. Graphical representation of B phase 7th harmonic voltage predictions.


8.000 7.000 Harmonic Distortion (Volts) 6.000 5.000 4.000 3.000 2.000 1.000 0.000 Predicted Voltage 7th Harmonic Simulated Voltage 7th Harmonic Step 2 Step 3 Step 4

Fig. 10. Graphical representation of C phase 7th harmonic voltage predictions.

Harm onic 5 7 11 13 17 19

Predicted Va (Volts) 3.922 7.374 0.999 0.615 0.303 0.245

Simulated Va (Volts) 3.960 7.558 0.981 0.601 0.306 0.242

Error -0.96% -2.43% 1.83% 2.33% -0.98% 1.24%

Predicted Vb (Volts) 5.202 6.650 1.075 0.790 0.365 0.308

Simulated Vb (Volts) 5.158 7.172 1.081 0.784 0.360 0.303

Error 0.85% -7.28% -0.56% 0.77% 1.39% 1.65%

Predicted Vc (Volts) 5.514 5.013 1.459 0.922 0.461 0.361

Simulated Vc (Volts) 5.498 5.371 1.459 0.901 0.460 0.360

Error 0.29% -6.67% 0.00% 2.33% 0.22% 0.28%

Table 4: Results of phase voltage predictions for the highest value of capacitance. 7. ACKNOWLEDGEMENTS The authors gratefully acknowledge the contributions made by Alstom T&D P&C Ltd to this research. 8. REFERENCES [1]. M.J.Bridgeman, R.E.Morrison, S.B.Tennakoon. Measurement of harmonic impedance on an LV system utilising capacitor switching and consequent predictions of capacitor induced harmonic distortion, paper, IEEE International Conference on Harmonics and Quality of Power (ICHQP) October 1998. [2]. EMTDC user manual Manitoba HVDC Research Centre. [3]. Hu, L and Morrison, R E. AC side equivalent circuit based method for harmonic analysis of a converter system, proc. IEE Electr. Power Appl., Vol 146, No 1 January 1999. 9. BIOGRAPHIES Mark Bridgeman was born in Bishops Stortford, England. He graduated from Southbank University and is currently working on research towards a PhD through Staffordshire University. His employment experience includes PFC Engineering and Alstom T&D P&C Ltd. His specialist field of interest is power quality issues. Mr Bridgeman is a Chartered Engineer and a member of the IEE in the UK. RE (Bob) Morrison was born in Stoke on trent, England in 1951. He graduated with a BSc in 1973 and a PhD in 1981. He worked with ALSTHOM from 1967 to 1983 and with Staffordshire University from 1983 until 1997. He is presently the Sir John Monash Professor of Electrical Power Engineering at Monash University, Victoria, Australia. His fields of interest include ac traction and power quality. Norman MacLeod was born in Dunbartonshire, UK in 1951. He graduated from the University of Strathclyde, UK in 1973 with a B.Sc. in Electrical Engineering and received a PhD from the same institution in 1979 for research into gas discharge physics. He joined the ALSTOM (then GEC) Power Transformer company in 1976 as a development engineer, transferring to the Power Electronics company in 1981 to work on program development and insulation co-ordination. He has worked on the fields of harmonic filter design, Static Var Compensators and HVDC and is now the manager of their Project Engineering department. Dr MacLeod is a Chartered Engineer and a member of the IEE in the UK, a member of the IEEE/PES and a member of CIGRE. Sarath Tennakoon was born in Maho, Sri Lanka on January 18th 1953. He received the BSc(Eng.) Hons degree in Electrical Engineering from University of Sri Lanka, M.Sc degree in Electrical Engineering from Aston University, Birmingham, U.K. and the PhD degree from CNAA, U.K. in 1975, 1980 and 1987 respectively. From 1985-1987, he was with ALSTOM Transmission and distribution Projects Ltd., working in the HVDC valve design group. In 1987 he joined the School of Engineering and Advanced Technology, Staffordshire University, U.K.. Dr Tennakoon is a Reader in Electrical Engineering with research interests in Power Electronic Applications in Power Systems.

Vous aimerez peut-être aussi