Distribution. Paper ABSTRACT CTION INTRODUC. from the network. electronics. with internal. likely caused by. resonances in. are.

Transcription

1 HAMOIC ESOACES I ESIDETIAL LOW VOLTAGE ETWOKS CAUSED BY COSUME ELECTOICS Jan MEYE, obert STIEGLE, Peter SCHEGE Technische Universität Dresden - Germany jan.meyer@tu-dresden.de Ingolf ÖDE, Andreas BELGE M etzdienste hein-main GmbH - Germany i.roeder@nrm-netzdienste.de ABSTACT Harmonic resonances in electricity networks can cause significant amplifications harmonic voltages/currents. They can also affect stable operation power electronics with internal stw controls, like phovoltaic inverters (harmonic instabilities). Up now capacir banks treated as main reason for resonances in residential low voltage networks as long as y detuned, no resonances at frequencies below 2 k expected. However, harmonic s in an urban residential network without dedicated capacir banks have indicatedd existencee a significant resonance in range 500. This paper presents resultss a detailed study network harmonic impedance in this low voltage network based on simulations. It shows that residential cusmers can introduce a significant amount distributed capacitance, which cannot be neglected. etwork harmonic impedance in or residential networks confirm general existencee such resonances. The paper proposes an updated model residential users for harmonic studies. ITODUC CTIO Public low voltage (LV) networks, particular those with a dominating sh residential cusmers, usually assumed be free resonances in harmonic range below 2 k, except in presence dedicated non- detuned capacir banks. This is mainly justified by small tal length lines traditional P/Q load model residential cusmers with inductive characteristics s. Consequently for harmonic studies or calculation harmonic emission limits (e.g. according D-A-CH-CZ rules for assessment network disturbances [] or VDE A- 405 [2]) ten only simplified impedance line model, whichh correspondss short circuit impedancee multiplied by harmonic order (h Z sc ), is considered. On or h, utilities report about an increasing number cases with elevated levels at harmonic orders between 0 20, sudden changes in harmonic voltages currents at se orders while switching larger installations (e.g. streett lighting) or significant amplifications mains signalling levels. These phenomena usually linked with resonances. Often growing sh capacitance in modern power electronic equipment is discussed have an increasing impact on network harmonic impedance in future (e.g. [3], [4]). Particular new energy efficient equipment operating at higher switching frequencies (e.g. phovoltaic inverters, electric vehicle chargers) requires grid-side filter circuits with larger capacitances in order keep switching frequency emission (supraharmonics) away from network. This paper presents a detailed study network harmonic impedance in an urban residential low voltage network with a distinct resonance around 500. Based on simulations it confirms that resonance is most likely caused by cusmer equipment. The paper starts with a brief overview typical resonance situations in public LV networks. After some details about studied network, network harmonic impedancee presented discussed. The next section introduces a harmonic model for households, which enables a realistic simulation measured network harmonic impedance. Finally some measures for reduction harmonic voltages briefly discussedd general recommendations with regard harmonics in LV networks with high sh modern power electronics provided. TYPICAL ESOACEE SITUATIOS In general, two different types resonances have be distinguished: series resonances parallel resonances. A series resonance can be observed on MV side is composed by series circuit MV/LV transformer reactance a capacitance at LV side (Fig. a)). A parallel resonance can be observed on LV side is composed by parallell circuit MV/LVV transformers reactance a capacitance at LV side (Fig. b)). A resonance itself is usually not a problem as long as it is not excited. The critical excitation for a series resonance is a harmonic voltage close resonance frequency, where impedance resonance circuit has a minimum. 20 kv a) 400 V M ~ U h Ih Upstream network 4000 V Fig. Typical resonance situations in LV networks [] a) series resonance; b) parallel resonance M ~ b) I h /5

2 T ( h ) U V j X h C Fig. 2 Equivalent circuit series resonance with critical excitation (X T transformer reactance; X V cusmer reactance; X C capacir bank; cusmer resistance) T Fig. 4 Phase angle input impedance different power electronic devices for mass-market V j X h C I ( h ) Fig. 3 Equivalent circuit parallel resonance with critical excitation (for symbols refer Fig. 2) Fig. 5 Phase angle input impedance a street light with reactive power compensation based on a shunt capacir Consequently, a high harmonic current flows, whichh in turn causes high harmonic voltages across individual elements circuit (Fig. 2). This way e.g. a mains signalling voltage injected at MV side can be amplified in a downstream LV network. The critical excitation a parallel resonance is a harmonic current close resonance frequency, where impedancee resonance circuit has a maximum (Fig. 3). Consequently, injected harmonic currents can significantly amplify harmonic voltages. Sources such harmonic currents can be e.g. current source converters. It should be noted that characteristicc dominant harmonic source plays an important role. In case voltage source behavior, a parallel resonance is not excited at all, as due high impedance flowing harmonic currents do reduce. Furrmore, in case harmonic resonances origin excitation location, where interference occurs, can differ significantly (e.g. damage a non-detuned capacir bank in an LV installation withoutt any disrting equipment, just due existing harmonic voltages in upstream MV network). In case a dedicated capacitance, like a capacir bank, resonance frequency f can be estimated by Ssc f = f () Qc The short-circuit power S sc is linked reactancee upstream grid, capacitive reactivee power Q c capacitance at considered location (f : power frequency). In case switched capacir banks resonance frequency switches as well hit a frequency, where a relevant excitation exists, is more likely. However resonance can easily be mitigated by detuning capacir bank, which is always recommended d. The issue becomes more complicated, if sum capacitances is distributedd throughout whole LV network. In this case eq. () provides less accurate results detuning is virtually not possible at. One origin such distributed capacitances input circuits power electronic equipment. Measurements input impedance mass-market equipment have shown that a dominating sh has in relevant frequency range negative phase angles, which means capacitive characteristic. Some examples presented in Fig. 4. Anor source distributed capacitances shunt capacirs for power facr correction inside devices (e.g. street light). Fig. 5 exemplarily presents phase angle input impedance a street light, which has a power facr about 0.95 ind at 50, but behaves purely capacitive already at second harmonic. This raises question, if equipment design must not only consider efficiency at power frequency, but also network- friendly impedance at harmonic frequencies. ETWOK UDE STUDY Fig. 6 on next page provides an overview studied network. This network has been selected, because routine System Operar M etzdienste hein-main GmbH have indicated existence a resonance welll below 2 k. In particular, an amplification mains signalling voltage by facr 2..5 sudden, significant changes in harmonic levels in case switching street lights have been observed. The network is located in Frankfurt/Mainn supplies about 240 housing units like single-family houses apartment houses as well as 85 street lights with characteristic presented in Fig. 5. The network has been built about 5 years ago on greenfield. This suggests that most households have a dominating sh newly bought energy-efficient power electronics, with a significant amount capacitance in grid (cf. Fig. 4). The network has 3 junctionn boxes is operated as meshed grid. Consequently, short circuit power is with about 3 MVA at junction boxes about same rar high. Usually in such strong, newly built networks no Power Quality problems expected. IMPEDACEE MEASUEMETS Measurement system The measurement system has been developed at Technische Universitaet Dresden is able measure single phase loop network harmonic impedances in frequency range between k. A scheme 2/5

3 Fig. 6 Schema studied network system is shown in Fig. 7. It consists a programmable amplifier that injects a current variable frequency in network. All taken with sampling rates up MS/s. The amplifier as well as measurement instrument controlled by a computer. In order avoid any impact measurement system itself on results, it is supplied by an emergency power supply unit. In order disturb network connected cusmer equipment as less as possible, a single frequency sweep with adaptive magnitude injected current has been performed. The network harmonic impedance calculation is based on differencee between harmonic voltages currents before during injection. To minimize impact background disrtion on results, interharmonic frequencies used for injection. In general, two different types network harmonic impedances calculated: average impedance determined based on multiple cycles at power frequency (common method) point-on-wavevariation within a cycle at power frequency. This impedance analyzing impedance variation is mainly caused by power electronic devices with simple rectifier bridges. During recharging state rectifier is conducting DC-lingrid. Consequently, input impedance device is highly capacitive. In case capacir is directly connected rectifier bridge is not conducting (no recharging) impedance device is usually very high. Furr details on measurement setup network harmonic impedance determination can be found in [5]. Measuremen nt results The network harmonic impedance has been measured at junction boxes (B, C, D in Fig. 6) LV busbar (A in Fig. 6). Fig. 8 exemplarily presents magnitude phase angle average impedance measured during day, when street lights switched f. Measurements at different times day did not show significant differences. At alll locations A-D a significant parallel resonance around 500 is observed. At frequencies below 800 similar at all three junctions boxes (B-D) higher compd measurement at LV busbar (A). In order analyze balance between phases, consecutive on all phases has been performed at each location. As example Fig. 9 presents se for junction box C. While resonancee frequency is almost similar, some differences exist in magnitude, particular between L3 L/L2, whichh might result from an unbalance in number type connected equipment between three phases. As frequency used for mains signalling is with 482 very close resonance frequency, it is confirmed that aforementioned amplification mains signalling levels in this LV grid is caused by resonance. In case street lights switched on, resonance shiftss slightly lower frequencies, because more capacitance is connected grid. Furr details presented discussed in next section. SIMULATIOS Alll simulationss have been performed in DigSilent PowerFacry using «Harmonics» module. etwork schema cusmer data (location, active reactive power) has been provided by utility. It exactly corresponds what has been used by utility during network planning check if voltage b is not violated Fig. 7 Harmonic network impedance measurement system (layout pho) 3/5

4 Fig. 8 etwork harmonic impedance at different locations (phase ) Fig. 9 etwork harmonic impedance at measurement location C network elements not overloaded. As first step network harmonic impedance has been simulated for initial network configuration. The results presented in Fig.. The plots correspond a line, which is determined by short circuit impedance at measurement location. Consequently, it is almost similar for junction boxes (B-D), but lower for LV busbar (A). o resonance exists at all, which confirms that line capacitances cannot be reason for measured resonance around 500. As tal amount power electronic equipment is expected be main reason for measured resonance, a model representing network harmonic impedance characteristic an individual household has been developed. Based on measured resonance frequency known short-circuit power () can be converted calculate equivalent capacitance: C = 2 π f U 2 f f To derive model several more or lesss complex circuit configurations has been compd. The circuit presented in Fig. 0a) has been finally identified as most suitable one. Simpler circuits usually not able reflect properly low resonance rise high resonance width. The model has been implemented for each household parameters have been adjusted iteratively until resulting network harmonic impedance has matched measured one as best as possible. The results 2 S sc (2) C 5 µf; C 0,5 Ω; L 95 mh; L 30 Ω C L 2 µf L L 470 mh; L 48 Ω a) household b) street light Fig. 0 Specific harmonic simulation models loads presented in Fig. 2 ger with measured values (dots). It should be noted that model provides satisfying results at frequencies between about While magnitude accuracy at junction boxes is very good (B-D), calculated network harmonic impedancee magnitude at busbar is slightly low. Like CIGE load model for harmonic studies in HV systems as described in [6], model proposed in this paper is also not suitable for power frequency load flow calculations. As input impedance power electronic equipment depends significantly on frequency (cf. Fig. 4), parameters equivalent circuit also frequency-dependent, which explains upper frequency limit for application. Finally, a frequency-dependent model for street lights has been addedd simulation in order study reasons for sudden changes in harmonic voltage levels while switching street lights. Based on lab frequency-dependent input impedance a single lamp, circuit as shown in Fig. 0b) has been developed. It has been implemented in Fig. etwork harmonic impedance simulated without load Fig. 2 etwork harmonic impedance without streetlights Fig. 3 etwork harmonic impedance with streetlights O (grey: streetlights OFF) 4/5

5 simulation using an equal distribution street lights three phases. The results presented in Fig. 3. Switching street lights O shifts resonance by about 50 wards lower frequencies. As consequence network harmonic impedance changes considerably for specific harmonics (e.g. 9,, 3, 5). Assuming a more or less constant harmonic current at se harmonic orders sudden change network harmonic impedance can explain sudden change in harmonic voltages that has been observed in s, particular for harmonic orders close resonance frequency. COMPAIS SO WITH OTHE GIDS In order check, if resonance observed in studied grid is a unique issue, three or grids with similar configuration have been measured compd with studied grid. The selected grids located in same wn like studied grid. While one grid comps in age with studied one or two grids older. As rationale for different ages serves assumption that in particular older grids might not have such a high penetration with modern power electronicc equipment than newly built ones. Table I comps results. Table I Measurement results from different LV grids Grid Frequency Magnitude first resonance impedance Studied grid (5 years old) 480 0,33 Ω 5 years old 750 0,3 Ω 5 years old ,25 0,3 Ω 30 years old 780 0,33 Ω All grids have a distinct resonance below k with comparable impedance magnitudes. However, resonance frequency or grids is still higher than in studied grid probability its excitation is consequently a bit lower. The results suggest that such low resonancee frequencies, like in studied grid, seem not be common in distribution a utility yet. Level amplification mains signalling voltages can be used as simple indicar for resonances around 500. Anyway it is recommended monir network harmonic impedance in selected LV grids in future identify possible shifts first resonance (magnitude frequency). COCLUSI IOS The paper presents a detailed study a harmonic resonance observed in an urban residential LV grid. The resonance is located around 500 (0th harmonic), which is very low commonly not expected in LV networks. Simulations have shown that resonance is most likely caused by capacitance introduced by sum power electronic equipment used in households. Particular grid side filter circuits in modern energy-efficient equipment operating hat high switching frequencies seem contribute significantly capacitance causing resonance. As se capacitances distributedd in network, detuning like in case dedicated capacir banks is no option. In case resonance causes harmonic voltages above limits (e.g. according E 5060), most promising option is reduction harmonic currents, e.g. by active filtering. Furr simulationss have shown that such a filter is most effective, if it is placed close one junction boxes if it controls injected harmonic current based on moniring harmonic voltages at junction boxes or feeder ends or based on a minimization tal harmonic current measured at LV busbar for considered harmonic orders. The presented study has clearly shown that harmonic resonances in LV networks cannot be neglected anymore. etwork operars should seriously think about considering such issues already in process network planning. As consequence results equations for calculating harmonic emission limits in ongoing revision relevant rules guidelines for assessment network disturbances (e.g. D-A-CH-CZ or VDE A- 405/400) have been extended by a resonance facr. It is also recommended monir network harmonic impedance in selected networks once in one or two years, in order identify possible trends first resonance. A harmonic load model for households based on a L C circuit has been proposed successful parametrized for studiedd network. However furr studies needed validate general applicability model. Along with this, more sophisticated procedures for parameter identification has be explored. ACKOWLEDGEMET The authors wish thank Henning Hauptmann, who developed major parts project in his Diploma sis. EFEECES [] D-A-CH-CZ Technical ules for Assessment etwork Disturbances, 2nd edition, [2] VDE A 405: Generars connected low-voltage distribution network - Technical requirements for connection parallel operation with low-voltage distribution networks, 20 (in German) [3] D. Chakravorty; J. Meyer; P. Schegner; S. Yanchenko; M. Schocke, "Impact Modern Electronic Equipment on Assessment etwork Harmonic Impedance," in IEEE Transactions on Smart Grid, vol.pp, no.99, pp.-.. [4] A. S. Koch, J. M. A. Myrzik, T. Wiesner L. Jendernalik, "Harmonics resonances in low voltage grid caused by compact fluorescent lamps," Proceedings 4th International Conference on Harmonics Quality Power - ICHQP 200, Bergamo, 200, pp. -6. [5]. Stiegler, J. Meyer, P. Schegner D. Chakravorty, "Measurement network harmonic impedance in presencee electronic equipment," 205 IEEEE International Workshop on Applied Measurements for Power Systems (AMPS), Aachen, 205, pp [6] CIGE working group 36-05, Harmonics, characteristic parameters, methods study, estimates existing values in network Electra journal, vol. 77, July 98. 5/5