Switching transients in wind farm grids Poul Sørensen1), Anca D. Hansen, Troels Sørensen2), Christian S. Nielsen2), Henny K. Nielsen3), Leif Christensen3), Morten Ulletved3) 1)
Risø National Laboratory, Technical University of Denmark, VEA-118, P.O.Box 49, DK-4000 Roskilde, Denmark, e-mail
[email protected], tel +46 4677 5075, fax +46 4677 5083. 2)
DONG Energy, Teglholmen, A.C. Meyers Vænge 9, DK - 2450 København SV, Denmark.
3)
DELTA Dansk Elektronik, Lys & Akustik, Erhvervsvej 2 A, DK-8653 Them, Denmark.
Summary This paper presents measurements of voltage and current transients performed in the medium voltage power collection grid of Nysted offshore wind farm. A number of switching events have been performed, and 3 phase voltages and currents have been measured synchronously with 2.5 MHz sampling frequency in three points on a selected line in the 33 kV power collection grid to verify the transient behaviour including the wave propagation through this line. Preliminary simulations have also been performed of the closing of the breaker in the transformer platform connecting the instrumented 33 kV line, and the simulation results are compared to the measurements. Keywords: Wind farms, switching transients, measurement and simulation.
1 Introduction Although the wind power development is still mainly based on land sites, a number of large offshore wind farms have been developed, and there are significant plans for further offshore wind power development, e.g. in Denmark, Germany, the Netherlands and United Kingdom. The development towards larger wind power installations such as offshore wind farms has increased the focus from TSO’s on how this influences the operation of the power system on the transmission system level. The main concern from TSO’s point of view seems to have been on the faultride-through capability of the wind turbines, in order to avoid sudden loss of substantial wind power due to a fault in the transmission system [1], [2]. Another issue is to enable control of active and reactive power during fault free operation as well [2], [3]. Lately, there has also become focus on the power fluctuations, which have been observed to be larger from offshore wind farms than from distributed wind turbines on land sites [4], [5]. However according to Liljestrand et.al. [6], challenges coming with increased wind farm size concern not only how to handle the transport of power to shore and the impact on the transmission system, but also how to design and build the internal sub-sea cable grid interconnecting the wind turbines, often referred to as the power collection grid or power collection system [7].
This paper deals with switching transients in the power collection system. Liljestrand et al. [6] studied switching transients in a large wind farm based on simulations with PSCAD/EMTDC. Sørensen et al. [8] have reported transient simulations using the Power Factory software from DIgSILENT and compared to measurements acquired with 10 kHz sampling frequency on the Hagesholm 6×2 MW wind turbine cluster. In the present paper, measurements with 2.5 MHz sampling frequency of switching transients on the Nysted 165 MW offshore wind farm is combined with simulations using Power Factory. Figure 1 shows the power collection grid of Nysted offshore wind farm. It consists of 72 wind turbines, each with 2.3 MW rated power. The power collection grid is built as a 33 kV sea cable grid. The 8 lines from the 8 rows of wind turbines are connected 4 by 4 on the transformer platform and transformed to 132 kV, which is the voltage level of the cable on shore.
A1 H1
A9 H9 Figure 1. Power collection grid of Nysted offshore wind farm.
2 Test and measurement setup A comprehensive test procedure has been executed on the Nysted offshore wind farm. This test procedure included opening and closing of the breakers on the transformer platform to the individual lines A-H, and of the switchgear in the individual wind turbines A1 and A9. During the test, measurements have been acquired on the 33 kV line A at the cable end on the transformer platform, at wind turbine A1, and at wind turbine A9.
This paper will only deal with one of the tests: the closing of the line breaker for line A on the transformer platform. This test is particularly interesting because it visualises how the voltage waves propagate with a fraction of the speed of light through the line where measurements were acquired. 3 phase voltages to ground and 3 phase currents are acquired synchronously in 3 points on line A. The first measurement point is on the platform on the wind farm side of the line breaker. The other two measurement points are in wind turbines A1 and A9 respectively. The measurement points in the wind turbines are at the transformer terminals, i.e. the measured current is the current to the wind turbines. Note that the applied current sign convention is positive current from the grid to the wind farm and to the wind turbines, i.e. consumption sign. The line-to-ground voltages are measured with capacitive voltage dividers, which are connected to the data logger through a high bandwidth amplifier developed by DELTA. The bandwidth (3 dB) of the total voltage measurement is 1Hz to 10 MHz. The line currents are measured with flexible current clamps. The bandwidth (3 dB) is 0.55 Hz to 3 MHz. The data logging is performed with 2.5 MHz sampling rate. The data logging in the 3 measurement points is synchronised using GPS based solution. The accuracy of the synchronisation is within one sample, i.e. better than 0.4 µs.
3 Simulation model The simulations are performed with Power Factory from DIgSILENT. Two different time series simulation methods are available in Power Factory. The first method simulates RMS values (amplitudes and angles) taking into account electromechanical transients. This method is intended for power system stability studies. The second method, which is applied in this paper, simulates instantaneous values taking into account electromagnetic transients as well. Instantaneous value simulations with Power Factory have been shown to agree very well with corresponding simulations using PSCAD/EMTDC [8]. This is a useful observation, because Power Factory uses built-in component models (e.g. line models) with hidden source code, whereas PSCAD simulations are done with open component models. However, the good agreement has only been shown for bandwidth up to a few kHz. The single line diagram of the applied simulation model is shown in Figure 2. Compared to the actual grid layout in Figure 1, it is seen that only the lines A-D are included in the simulation model, while the lines E-H are omitted because they are connected to another transformer winding set than line A, where the breaker is closed. The transformer windings in the 33 kV grid are deltacoupled for the platform transformer as well as for the wind turbine transformers. Thus the 33 kV grid is floating, and it 132 kV network ~ V
Land cable
Radsted/B1
Tr.69 A9/B1
A9
Line B9/B1
Line C9/B1
Line D1D2 Line D2D3 Line D3D4 Line D4D5 Line D5D6 Line D7/B1
Tr.69 C8/B1
Tr.69 C9/B1
Line D9/B1
Platform Tr33b/B1
Platform Term 33b/B1
Platform Term 33a/B1
Transformer D6
Transformer D7
Tr.69 D7/B1
Line D8/B1
Transformer C9
Transformer D5
Tr.69 D6/B1
Line D6D7
Line D6/B1
Transformer C8
Transformer D4
Tr.69 D5/B1
Line D7D8
Line C1C2 Line C2C3 Line C3C4
Tr.69 C7/B1
Line C8/B1
Transformer B9
Tr.69 B9/B1
Line D0D1
Line C0C1 Tr.69 B8/B1
Tr.69 C6/B1
Line C7/B1
Transformer B8
Line D5/B1
Transformer C7
Transformer D3
Tr.69 D4/B1
Transformer D8
Tr.69 D8/B1
Line D8D9
Line B7B8 Line B8/B1
Transformer A9
Line C6/B1
Transformer B7
Tr.69 B7/B1
Line B8B9
Tr.69 A8/B1
Line A8A9
Line A8/B1
Line B7/B1
Transformer A8
Tr.69 C5/B1
Transformer C6
Platform
Tr.69 D3/B1
Line D4/B1
Transformer C5
Transformer D2
Tr.69 D2/B1
Line D3/B1
Transformer C4
Tr.69 C4/B1
Line C5/B1
Transformer B6
Tr.69 B6/B1
Line B6B7
Tr.69 A7/B1
Line A7A8
Line A7/B1
Transformer B5
Tr.69 B5/B1
Line B6/B1
Transformer A7
Line C4/B1
Line C4C5
Line B4B5 Line B5/B1
Transformer A6
Tr.69 A6/B1
Line A6A7
Line A6/B1
Tr.69 B4/B1
Line B5B6
Tr.69 A5/B1
Line A5A6
Line A5/B1
Transformer B4
Platform cable b
Tr.69 D1/B1
Line D2/B1
Transformer C3
Tr.69 C3/B1
Line C5C6
Line B4/B1
Transformer A5
Line C3/B1
Transformer D1
Line D1/B1
Transformer C2
Tr.69 C2/B1
Line C6C7
Line B1B2
Transformer A4
Tr.69 C1/B1
Line C2/B1
Transformer B3
Tr.69 B3/B1
Line B3B4
Line B3/B1
Transformer C1
Line C1/B1
Transformer B2
Tr.69 B2/B1
Line B2B3
Transformer A3
Tr.69 A4/B1
Line A4A5
Line A4/B1
Tr.69 B1/B1
Line B2/B1
Tr.69 A3/B1
Line A3A4
Line A3/B1
Line A9/B1
Transformer A2
Tr.69 A2/B1
Line A2A3
Line A2/B1
Transformer B1
Line B1/B1
Line C7C8
Tr.69 A1/B1
Line A1A2
Line A1/B1
A1
Line C8C9
Transformer A1
Line B0B1
Line A0A1
Platform cable a
Platform Tr132/B1 132/33/33 Transformer
Platform Tr33a/B1
Sea cable
Shore/B1
Transformer D9
Tr.69 D9/B1
Figure 2. Single line diagram of simulation model. Breaker and measurement position on transformer platform, and measurement positions on wind turbines A1 and A9.
is only coupled indirectly to ground through the capacitance to ground in the 33 kV cables.
The idea with the constant parameter model is first to find the resonance frequency using lumped models for the cables (typically some kHz), and then use this resonance frequency to calculate a set of distributed parameters that fits the travel time. With this option, the number of differential equations to solve is kept down, and the simulation time is almost the same as with lumped cable models. The frequency dependent parameter model fits the line parameters for a specified frequency interval, default from 1 mHz to 1 MHz. Such a fit results in a significant increase in the number of differential equations, and the simulation time gets significantly slower. In order to correspond to the sampling time of the measurements, the Power Factory simulations are done with 0.2 µs integration step size and an output step size of 0.4 µs, corresponding to 2.5 MHz.
Ap voltage [pu]
1 0 -1 -2 -0.1
0
0.1
0.2 time [ms]
0.3
0.4
0
0.1
0.2 time [ms]
0.3
0.4
0
0.1
0.2 time [ms]
0.3
0.4
2 A1 voltage [pu]
Different line model options are available. The standard lumped (π) parameter model is default, but also distributed line parameter models can be applied. There are two options for distributed parameter models: a constant parameter model and a frequency dependent parameter model.
2
1 0 -1 -2 -0.1
2 A9 voltage [pu]
The transformer models in Power Factory are standard Tequivalents, except that it is possible to include capacitance between the windings (HV-ground, LV-ground, HV-LV positive sequence and HV-LV negative sequence).
1 0 -1 -2 -0.1
Figure 3. Measured voltages in p.u. with base corresponding to 33 kV line-to-line rms.
4 Measurement results
The travel time from closing of phase 2 on the platform (t=0) till it reaches A9 is measured to 46 µs. With a cable length of 7061 m, the travel speed has been calculated to 0.51⋅c, where c is the speed of light. Figure 4 shows the measured currents in the same measurement points. The currents are calculated in p.u. using rated wind turbine current corresponding to 2.5 MVA. Thus, the p.u. base current is 9 times higher for Ap current than for A1 current and A9 current. The green spikes that occur in the wind turbine currents at A1 and A9 confirms the arrival time of the voltage wave.
Ap current [pu]
1 0 -1 -2 -0.1
0
0.1
0.2 time [ms]
0.3
0.4
0
0.1
0.2 time [ms]
0.3
0.4
0
0.1
0.2 time [ms]
0.3
0.4
1 A1 current [pu]
First it is observed that the selected sampling frequency is sufficient to reveal that the phases are not connected simultaneously. This is due to the different voltages over the circuit breakers in different phases. Although the voltage on the grid side is not measured, the expected voltage has been calculated based on the phase angle after the connection is completed. This has confirmed that the phase with highest voltage (the green phase 2 in this case) over the circuit breaker is also the phase where the breaker closes first. Liljestrand et.al. [6] included the voltage withstand capability of the circuit breakers in their model, and simulated that a pre-strike occurs when the voltage across the circuit breaker exceeds the voltage withstand capability.
2
0.5 0 -0.5 -1 -0.1
1 A9 current [pu]
Figure 3 shows the measured line-to-ground voltages on platform (Ap), wind turbine A1 and wind turbine A9. The voltages are calculated in p.u. using 33 kV line-to-line rms as base. Standard colours are selected for the phases: phase 1 is red, phase 2 is green (standard yellow), and phase 3 is blue.
0.5 0 -0.5 -1 -0.1
Figure 4. Measured currents in p.u. with base corresponding to number of wind turbines times rated wind turbine current.
5 Preliminary simulations
Figure 5 and Figure 6 shows Power Factory simulations of voltages and currents using lumped cable models for all cables in the system. As expected, this model simulation fits poorly with the measurements. Especially the simulated current spikes at breaker closing are dramatically, because half of the cable capacitance is lumped at the switching point.
Ap current [pu]
30 20 10 0 -10 -0.1
0
0.1
0.2 time [ms]
0.3
0.4
0
0.1
0.2 time [ms]
0.3
0.4
0
0.1
0.2 time [ms]
0.3
0.4
1 A1 current [pu]
In this chapter, Power Factory simulations are performed with lumped cable models and two different distributed cable models. Besides, the effect of simulating a capacitance to ground on the wind turbine transformer HV-windings is illustrated.
0.5 0 -0.5 -1 -0.1
1
1 0 -1 -2 -0.1
0
0.1
0.2 time [ms]
0.3
0.4
A9 current [pu]
Ap voltage [pu]
2
0.5 0 -0.5 -1 -0.1
A1 voltage [pu]
2 1 0
Figure 6. Power Factory simulation of currents with lumped cable model.
-1 -2 -0.1
0
0.1
0.2 time [ms]
0.3
0.4
0.05 0.04
A9 voltage [pu]
2 0.03
1 0
0.02
-1
0.01
-2 -0.1
0
0.1
0.2 time [ms]
0.3
0.4
Figure 5. Power Factory simulation of voltages with lumped cable model. Still, the simulation with the lumped model can be useful as a first step of the analysis using the distributed line parameter model with constant parameters. The frequency for fitting is first calculated by making an FFT on the simulations with the lumped model to identify the resonance frequency, and then calculate parameters that fit the travel time for this frequency. Figure 7 shows the FFT plot for the first line period of phase 1 after the breaker has closed on all phases. From that plot, the resonance frequency f = 1950 Hz is selected. Figure 8 and Figure 9 shows Power Factory simulations of voltages and currents using the distributed cable models with constant parameters (1950 Hz) for all cables in the simulated 33 kV grid. As expected, this model simulation agrees much better with the measurements than the lumped model. Still, there are significant differences. First of all, the voltages on all phases change at time 0 when the breaker in phase 2 is closed, while on the measurements, phase 1 and phase 3 voltages are kept very close to zero until the breaker closes on those phases.
-0.00 0
2
4
6
8
[kHz] 10
Figure 7. The Power Factory FFT calculation of the voltage in phase 1 in the first period after breaker closing. The reason for this difference is not yet clear, except that it is expected to be related to the representation of grounding and zero sequence in the simulation. The medium voltage grid is floating, as the transformers are delta-coupled and only the cable shields are grounded. Based on the wave travel time from platform to A9 observed in these simulations, the travel speed has been calculated to 0.38⋅c. Another significant difference between measurements and simulations is that current spikes are measured when the voltage waves arrive at the wind turbines, but these spikes are not visible in the simulations. This probably has to do with the capacitance between the wind turbine transformer HV windings and the ground. This will be illustrated later. Before this, simulations with the distributed frequency dependent parameter model for the cables are presented in Figure 10 and Figure 11. The distributed, frequency dependent parameter model is fitted to all frequencies f in the interval 1mHz < f < 1MHz. It takes significantly longer time to run this simulation than
the simulation with cable model fitted only to 1950 Hz, and the difference in the simulation result is only marginal. 2 Ap voltage [pu]
Ap voltage [pu]
2 1 0 -1 -2 -0.1
0
0.1
0.2 time [ms]
0.3
A1 voltage [pu]
A1 voltage [pu]
0 -1
0.1
0.2 time [ms]
0.3
A9 voltage [pu]
-1
0.4
0
0.1
0.2 time [ms]
0.3
0.4
0
0.1
0.2 time [ms]
0.3
0.4
0 -1
1 0 -1 -2 -0.1
0.2 time [ms]
0.3
2
0
0.1
0.2 time [ms]
0.4
1
0
0.1
1
-2 -0.1 0
0
2
2 A9 voltage [pu]
-1
0.4
1
-2 -0.1
0
-2 -0.1
2
-2 -0.1
1
0.3
0.4
Figure 8. Power Factory simulation of voltages with distributed constant parameter cable model (1950 Hz).
Figure 10. Power Factory simulation of voltages with distributed cable model fitted to all frequencies 1mHz < f < 1MHz.
. 2 Ap current [pu]
Ap current [pu]
2 1 0 -1 -2 -0.1
0
0.1
0.2 time [ms]
0.3
0 -0.5
0
0.1
0.2 time [ms]
0.3
0
0.1
0.2 time [ms]
0.3
0.4
0
0.1
0.2 time [ms]
0.3
0.4
0
0.1
0.2 time [ms]
0.3
0.4
0.5 0 -0.5 -1 -0.1
0.4
1 A9 current [pu]
1 A9 current [pu]
-1
1 A1 current [pu]
A1 current [pu]
1
0.5 0 -0.5 -1 -0.1
0
-2 -0.1
0.4
0.5
-1 -0.1
1
0
0.1
0.2 time [ms]
0.3
0.4
Figure 9. Power Factory simulation of currents with distributed constant parameter cable model (1950 Hz).
0.5 0 -0.5 -1 -0.1
Figure 11. Power Factory simulation of currents with distributed cable model fitted to all frequencies 1mHz < f < 1MHz.
Finally, Figure 12 shows simulated currents with distributed cable model fitted to all frequencies 1mHz < f < 1MHz, and with 0.5 nF capacitance on wind turbine transformer HVground. It is seen that the capacitances definitely causes current spikes in the simulations, but the size and frequency of these spikes is quite different from the measurements. This difference can, however, be related to the differences in the measured and simulated voltage waves.
Ap current [pu]
2
0 -1
0
0.1
0.2 time [ms]
0.3
0.4
A1 current [pu]
1
This paper describes results obtained in the project titled “Voltage conditions and transient phenomena in medium voltage grids of modern wind farms”, contract 2005-2-6345 funded by Energinet.dk, and it is carried out in cooperation between DELTA (project manager), DONG Energy, Vattenfall and Risø National Laboratory. Vattenfall is also acknowledged for their present participation in the project.
0.5
References
0 -0.5 -1 -0.1
0
0.1
0.2 time [ms]
0.3
0.4
1 A9 current [pu]
The measured current peaks are not present in preliminary simulations due to the simplified wind turbine transformer model, but current spikes are simulated when using a small capacitor (0.5 nF) to represent the capacitance between transformer HV and ground. These current spikes are very sensitive to this parameter, which therefore needs to be known more precisely in more detailed studies.
Acknowledgement
1
-2 -0.1
The preliminary simulations overestimate the travel times of the voltage waves.
0.5 0 -0.5 -1 -0.1
0
0.1
0.2 time [ms]
0.3
0.4
Figure 12. Power Factory simulation of currents with distributed cable model fitted to all frequencies 1mHz < f < 1MHz, and with 0.5 nF capacitance on wind turbine transformer HV-ground.
6 Conclusion Preliminary simulations are presented using available data for cables and transformers. As expected, the lumped cable models are not sufficient for this kind of transient simulations. Simulations are made with two different types of distributed parameter models: constant parameter and frequency dependent parameters. The results are quite similar. The constant parameter model simulates much faster, but it requires selection of the frequency for which the parameters are calculated. Thus, it may be easier to apply the frequency dependent parameters. The simulations with distributed parameters have similarities with measurements, but also significant differences. The main difference is that the voltage changes in all phases in the simulation when only one phase breaker closes, while in the measurements, only the phase where the breaker closes changes the voltage. The reasons for this are still unclear, although it must be related to the representation of the zero sequence, including the grounding of cable shields.
[1] C. Jauch, P. Sørensen, B. Bak-Jensen) International review of grid connection requirements for wind turbines. Nordic wind Power Conference NWPC 2004, Chalmers University of Technology. [2] Wind turbines connected to grids with voltages above 100 kV - Technical regulations for the properties and the control of wind turbines,, Energinet.dk, Transmission System Operator of Denmark for Natural Gas and Electricity, Technical Regulations TF 3.2.5, 2004, 35 p. Available: www.energinet.dk [3] J. R. Kristoffersen, P. Christiansen. Horns Rev offshore wind farm: its main controller and remote control system. Wind Engineering Volume 27, No 5 2003, p 351-360. [4] V. Akhmatov, J. P. Kjaergaard, H. Abildgaard, “Announcement of the large offshore wind farm Horns Rev B and experience from prior projects in Denmark”, European Wind Energy Conference, EWEC 2004. London. November 2004. [5] P. Sørensen, N. A. Cutululis, J. Hjerrild, L. Jensen, M. Donovan, L. E. A. Christensen, H. Madsen, A. Vigueras-Rodríguez. Power Fluctuations from Large Offshore Wind Farms. Nordic wind Power Conference NWPC 2006, Helsinki. [6] L. Liljestrand, A. Sannino, H. Breder, S. Johansson. Transients in Collection Grids of Large Offshore Wind Parks. NORDIC WIND POWER CONFERENCE, 2223 MAY, 2006, ESPOO, FINLAND [7] IEC 61400-21. Measurement and assessment of power quality characteristics of grid connected wind turbines. First edition. Dec 2001. [8] P. Sørensen, A. D. Hansen, P. Christensen, M. Mieritz, Jk Bech, B. Bak-Jensen, H. Nielsen. Simulation and verification of transient events in large wind power installations. Risø R-1331. October 2003.
