EVALUATION OF FLICKER LEVEL IN A T&D NETWORK WITH A LARGE AMOUNT OF DISPERSED WINDMILLS

Transcription

1 EVALUATION OF FLICKER LEVEL IN A T&D NETWORK WITH A LARGE AMOUNT OF DISPERSED WINDMILLS H. Knudsen V. Akhmatov The amount of installed wind power capacity in Denmark is very large, when seen relatively to the consumption. According to the prognoses up to the year 2005 a significant increase is expected in the southern part of Eastern Denmark. There is expected about 300 MW of the installed wind power capacity on-land in the region. The network in the region is relatively weak. The on-land windmills are connected to the grid at the MV distribution radials. The increasing amount of windmills in the MV distribution network generates a number of technical problems both in the MV distribution network and in the HV transmission network, and one of these problems is flicker. The existing methods of the flicker evaluations are statistical and only suited for radials at one MV level. With the increasing amount of windmills the problem concerning flicker is extending itself into the higher voltage levels from which they can penetrate back into other MV distribution networks. By this way, the flicker from windmills tends to be a system-wide problem, which is present both in the HV transmission network and in the local MV distribution network, as illustrated in Fig. 1. The flicker becomes the result of superposition of the voltage disturbances coming from different MV and HV levels so that the system-wide problem, where the existing methods of the flicker evaluations cannot be used. It is, therefore, necessary to develop suitable methods of flicker evaluations at these conditions. The method of flicker evaluation, developed in this work, is based on time-domain simulations using large network models with many scattered grid-connected windmills. The windmills are the sources of the disturbances imposing the power system. The semi-empirical windmill model is set up in the flicker range of 0.1 to 10 Hz and the windmill model is applicable with dynamic stability programs. Use of the dynamic stability tools provides an accurate network representation. In the simulations, the voltages in the chosen nodes are observed and post-processed where from the flicker levels will be found. The developed method treats penetration of flicker from the local MV network to the HV transmission network and back to other local MV distribution networks and superposition of flicker in the T&D network. The introduced method is treated on the physical model of the Eastern Danish power system with more than 60 windmill on-land settings. It is found that the flicker levels may be close to the flicker limit in the 10 kv distribution networks only and only in the worst case situation. The flicker levels are not found to be critical at the higher voltage levels. Fig. 1. The network model (simplified) and the flicker penetration throughout the network.

2 EVALUATION OF FLICKER LEVEL IN A T&D NETWORK WITH A LARGE AMOUNT OF DISPERSED WINDMILLS H. Knudsen V. Akhmatov 1. INTRODUCTION The amount of installed wind power capacity in Denmark is very large, when seen relatively to the consumption. So far the majority of windmills are landbased, but in the future also large offshore wind farms will be taken into operation. The land-based windmills are dispersed in the landscape or collected in several small and medium-sized wind farms. The on-land windmills are connected to the grid at the MV distribution radials. In the long run the major development will most likely be in the offshore wind farms, but in the nearest future one may still expect a significant number of new land-based windmills. According to the prognoses up to the year 2002 a significant increase is expected in the southern part of Eastern Denmark. On the major islands in this region, there is expected up to about 300 MW of the installed wind power capacity on-land in the year 2005 [1]. The network in the region is relatively weak, so problems are expected to arise because of that. 2. FLICKER PROBLEM FORMULATION The increasing amount of windmills in the MV distribution network on land will generate a number of technical problems both in the MV distribution network and in the HV transmission network. One of these problems is flicker caused by the varying power output from the windmill [2], which together with the subsequent changes in reactive power causes voltage variations in the T&D network. When the amount of installed wind power is small these problems are, in general, confined to the MV radial distribution network at which the windmills are connected. With the increasing amount of windmills the problem is extending itself into the higher voltage levels from which they can penetrate back into other MV distribution networks. By this way, the flicker from windmills tends to be a system wide problem, which is present both in the HV transmission network and in the local MV distribution network, as illustrated in Fig. 1. In order to investigate such problems it becomes necessary to perform larger system studies. The existing methods of flicker investigations are statistical and only suited for radials at one MV level. They do not take into consideration that the flicker level in a point may be the result of superposition of several flicker sources in various different locations in the network. For that reason they cannot be used for system-wide investigations. The possible solution can be time-domain simulations with use of large network models with windmills where the power system is imposed disturbances from the windmills. In the simulations, the voltages in the chosen nodes throughout the power system are observed and post-processed where from the flicker levels are found. Fig. 1. The network model (simplified) and the flicker penetration throughout the network.

3 In such investigations, the choice will be between transient programs and dynamic stability programs. The dynamic stability programs are the obvious choice because: 1. In the dynamic stability programs it is easy to make large network models with sufficient accuracy and reasonable effort (such network models may even be available beforehand). 2. The dynamic stability programs are only valid for low-frequency variations and the disturbances from windmills in normal operation are exactly in the low-frequency range [2,3]. 3. The dynamic stability programs are faster to run in case of large network models because they use the phasor representation, which allows larger time steps than the transient programs with instantaneous phase values. 3. NETWORK MODELLING Investigations of the flicker is made using a physical model of the 50 kv 132 kv 400 kv power system of the Eastern Denmark. The model is implemented in the dynamic simulation tool PSS/E and with the consumption and the power generation in thermal power plants (the northern part of the power system), a large number of decentralised combined heat-power units, and a large amount of land-based wind power. The windmill settings are scattered throughout the power system. In the model of the power system they are represented by more than 60 on-land windmill equivalents where the largest equivalent is of 30 MW of installed power capacity and the smallest one is less than 1 MW. Each equivalent has a size equal to the sum of all the windmills being connected to the given point of the 50 kv network. The weather conditions in the region is assumed to be strong wind, so that the power production from the windmills is the highest possible, which - from experience is 90% of the installed power capacity. This is the worst case with respect to the flicker level. The windmills are equipped with induction generators with the rated voltage of 0.7 kv and they are no-load compensated. Via the 0.7 kv / 10 kv transformers and, then, 10 kv / 50 kv transformers they are connected to the 50 kv local distribution network, as shown in Fig WINDMILL REPRESENTATION The power variations from the grid-connected windmills are simulated using the dynamic model representing disturbances of the mechanical power of the windmills. The model is developed by a semi-empirical way [2,3] by analysis of the measured electric power from a Danish wind farm and its Fourier-spectrum, see Fig. 2. The frequency range of the dynamic model is from 0.1 Hz up to 10 Hz. The dynamic model of windmills [2,3] connects the disturbances of the mechanical power being in the frequency range of the dynamic model with the excitation of eigenswings in the windmill mechanical construction. The windmill model is implemented in the dynamic tool PSS/E and evaluated by simulations. The simulation results, shown in Fig. 3, are found to be in agreement with the measurements [2,3]. 5. FLICKER COMPUTATION METHOD The flicker computation method is based on the following considerations. The grid-connected wind turbines are the sources of the disturbances. Excited in the windmill mechanical construction, the disturbances will be registered at the windmill induction generator terminals as voltage disturbances flicker. Fig. 2. (a) Measured electric power, (b) and its Fourier spectrum with marking: 1.-the 1 st harmonic of mill asymmetry, 2. and 3.- the 1 st and the 2 nd harmonics of vortex-tower-interaction, respectively, 4. -shaft torsion, 5. -blade swings (presumably), 6. -gearbox swings and mismatches (presumably).

4 Fig. 3. (a) The simulated active, electric power of the wind farm, (b) its Fourier spectrum. From there the flicker will penetrate to the local MV distribution network, and the flicker proceeds its penetration to the HV transmission network. Throughout the transmission network, the flicker will be, further, transmitted to the other local MV distribution systems, as shown in Fig.1, and may result in the voltage disturbances there. By this way, the flicker may be seen in the local MV networks, even where originally there are no disturbance sources, and there will be superposition of flicker. At the computation start, the following assumptions are made. The wind is irregularly distributed over the power system area why the operational points of the windmills in the different on-land settings are also different. The electric power in p.u. and the voltages of the windmills in the different on-land settings will be, then, different. The slips of the windmill induction generators and, then, the mill rotational speeds are also a little different from one setting to another. The electric link between the scattered windmill induction generators is negligibly weak and, hence, no (long-term) synchronisation of the windmills throughout the power system will be expected. The computation is started where the windmills have different blade-totower positions randomly distributed throughout the power system so that the windmills are out-of-phase. The worst case with respect to flicker would be obtained when all the windmills are at their rated operation and rotating in-phase (synchronous operation). That situation is, however, not realistic in the large power systems by the above-mentioned reasons. 6. DYNAMIC SIMULATIONS The computation proceeds dynamically in the time being necessary for making estimation of the flicker level in the chosen nodes. In accordance with the method described above, a number of simulations are made. The voltage dynamic behaviours at different voltage levels are shown in Fig. 4. The largest voltage fluctuations are found in the MV distribution network where the flicker sources are electrically closest. The voltage fluctuation magnitudes are attenuated in the HV transmission system. In all the voltage behaviours the beating phenomenon is observed. The beating is the result of mutual deviation between the wind turbine rotational speeds. Some wind turbines are rotating a little faster than other ones why the relatively large groups of the windmills can be rotating in-phase during short time periods temporary synchronisation, and after awhile they will be rotating out-of-phase again. The voltage fluctuations are largest when the windmills are rotating in-phase. 7. FLICKER EVALUATION From the simulated voltage vs. time series the flicker level can be evaluated in the chosen nodes in the power system. In [4] a functional description of a flickermeter is given. Based on the functional description of a flickermeter and knowing the flicker limit at various low-frequency disturbances it is relatively straight-forward to estimate the flicker level from the simulated RMS voltages, when one has performed an FFT analysis of the voltages and determined the relative amplitudes of all the lowfrequency variations in the signal. This is the definition of the short-term flicker level, P ST, where the measuring period is 10 min. The long-term flicker level, P LT, is defined, in accordance with [5], from the 12 consecutive P ST 12 flicker measurements 1 3 P = 3, 12 P. The LT ST i i= 1 measuring period of P LT is, then, 2 hours.

5 1. The flicker is computed using the windmill model where the mechanical disturbance pattern is not changed during the computation time. 2. The values of P ST found in the same simulation and in the same node are very close to each other and, therefore, to their value of P LT. From this point of view, the limit P LT =0.25 [5], shall be chosen. The flicker levels found in the simulations are collected in Tab. 1. The results are in case of a strong wind where the power production from the windmills is around 90%. Tab. 1. Computed flicker levels throughout the T&D network. The flicker is only caused by the windmills. Network 10 kv 50 kv 132 kv 400 kv P ST As can be seen, the highest flicker level is found at 10 kv level in the distribution networks. The flicker levels there are in some nodes close to the flicker limit of In other parts of the MV and the HV network the flicker levels are below the limit. 8. CONCLUSIONS Fig. 4. Simulated RMS voltages in: (a) the MV distribution network at 10 kv -level, (b) the 50 kv distribution network, (c) in the 132 kv transmission network, (d) in the 400 kv transmission network. In the power system, the flicker is caused by many different sources. The introduced method for flicker evaluation is only taking into consideration the flicker regularly coming from the windmills in normal operation why it is necessary to apply the acceptable flicker limit where a margin with respect to other possible flicker sources is taking into account. In [5] there are defined two flicker limits that are 0.35 for P ST and 0.25 for P LT, respectively. Both limits are defined for flicker in the MV and the HV networks and applied to windmills only. It is necessary to choose the limit from these two possible that will be relevant for the simulated case. This can be done using the following considerations: The flicker evaluation method has been introduced where the flicker levels in the nodes throughout the given T&D network are found from the simulated voltages. The flicker evaluation method relates to large power systems with many scattered flicker sources in this case to a large number of windmill on-land settings. The introduced method is based on that the disturbances excited in the windmill mechanical constructions are transferred to the windmill generator electric terminals where they are seen as the voltage fluctuations flicker. The windmill model is set up with the range of 0.1 to 10 Hz why the model is applicable in dynamic stability programs. Use of the dynamic stability tools provides an accurate network representation. From the windmill terminals the flicker is transferred to the local MV distribution network and, then, to the HV transmission network. Via the transmission network, the flicker can be, further, transferred to the other local MV distribution networks with superposition of flicker. By this way it is possible to evaluate the flicker levels in the chosen nodes throughout the T&D network, even in the HV transmission networks and the MV local distribution networks without any flicker sources in there what is not available from conventional routines for flicker calculations.

6 The introduced method has been used on the physical model of the Eastern Danish power system with more than 60 windmill on-land settings. It is found that the flicker levels may be close to the flicker limit only in a few nodes in the 10 kv distribution networks and only in the worst case situation with maximum wind power production. The flicker levels are not found to be critical at the higher voltage levels. REFERENCES 1. Bruntt M., Havsager J., Knudsen H., 1999, Incorporation of wind power in the East Danish power system, Paper BPT , Proceedings IEEE Pow Tech 99 Conf, Budapest, Hungary, Aug.29-Sept Akhmatov V., Knudsen H., 1999, Dynamic modelling of windmills, Proceedings Int Conf on Pow Syst Trans IPST 99, Budapest, Hungary, June 20-24, 1999, Akhmatov V., Knudsen H., Nielsen A. H., 2000, Advanced simulation of windmills in the electric power supply, Int J of Elec Pow and Energy Syst, vol.22, no.6, International Electrotechnical Commission, IEC Report IEC 868, Flickermeter - Functional and Design Specifications, International Electrotechnical Commission, IEC Report IEC , Electromagnetic compability (EMC) Part 3: Limits Section 7: Assessment of emission limits for fluctuating loads in MV and HV power systems, 1996.