Presented in AUPEC 7, Pert, Western Australia, 9- December, 7 ANALYSIS OF HARMONIC DISTORTION LEVELS ON A DISTRIBUTION NETWORK Glenn Nicolson - Manukau Institute of Tecnology, Auckland, New Zealand Professor Vic Gosbell, University of Wollongong, Australia Asok Parsotam, Vector Ltd, Auckland, New Zealand Abstract It is generally acknowledged tat armonic distortion levels on distribution networks are slowly but steadily increasing due to te proliferation of non-linear loads. Wile customer complaints relating to armonic distortion levels are relatively uncommon, and measured levels of armonic distortion are typically well witin te limits specified by national regulations and international standards, te increasing trend may mean tat more active management by utilities will soon be required. Effective management of armonic distortion levels will require knowledge not only of te level of disturbance, but also of ow te disturbance levels vary wit time, location, and te connected load. Accepted metods of assessing armonic levels tat rely on (typically 95%) cumulative probability statistics are inadequate for tis level of insigt. In tis paper voltage total armonic distortion levels (THD v ) from a distribution network power quality survey are analysed in terms of variation wit time, te relationsip of te measured THD v levels to system load and te pysical caracteristics of te monitored sites.. Introduction Harmonic distortion levels are seldom te source of complaints from network customers. After issues of network reliability, most customer complaints arise from discrete power quality disturbances suc as sort-term interruptions, voltage sags or swells, and voltage fluctuations tat cause ligt flicker. Wile armonic distortion may not cause suc immediate and easily-observed impacts, it can cause some equipment to malfunction, and does result in additional power losses in bot customer and network equipment []. It is generally accepted tat te proliferation of nonlinear loads is leading to a slow but steady increase in armonic distortion levels on networks. Even if present levels of armonic distortion are considered to be well witin acceptable levels, tis situation may well cange if te increasing trend continues. Management of armonic distortion levels will be assisted by knowledge of ow distortion levels vary bot in place and in time. Use of cumulative probability statistics calculated over te survey period do not give sufficient information for tis purpose. Tis paper will describe te analysis of armonic distortion levels recorded as part of a continuous power quality survey on a New Zealand distribution network. Measured levels of voltage total armonic distortion ave been analysed wit regard to bot variation over time and also te load caracteristics of te monitored. Reasons for tis study Te data on wic tis paper is based is taken from a continuous PQ monitoring programme carried out by Vector (NZ) Ltd. Vector ave been actively involved in routine PQ monitoring since 999 and now ave over 3 PQ monitors installed in te Auckland network, including 3 monitors connected to te kv bus in zone substations. It is te data from tese zone substations tat is analysed in tis paper. An earlier study [] analysed variation in THD v levels across te 3 monitored sites. For te purpose of tis study, te sites were grouped according to weter te predominant nature of te connected load was commercial, industrial or residential. Te survey consisted of two commercial sites, eigt industrial sites, and tree residential sites. 95% cumulative probability statistics were calculated for te purpose of reporting te THD v levels at eac It was found tat THD v levels were lowest at commercial sites, followed by industrial sites, wit residential sites igest. Tis seemed counter-intuitive, as it could be expected tat industrial sites would ave a muc iger proportion of non-linear loads tan residential sites. One of te outcomes of te earlier study was identifying te need for furter investigation into armonic distortion levels on te network. 3 Analysis of THD v data In tis paper, THD v data from te 3 monitored sites as been analysed, covering te period of te mont of June 4. Te objective of te analysis was to identify any time-varying trends in THD v levels, and also to identify any relationsip between te THD v and te predominant load type and load magnitude. Te steps in analysing te THD v data for eac site were: 86
Presented in AUPEC 7, Pert, Western Australia, 9- December, 7 Plot te THD v measurements against time in order to identify any trends wit regard to time (ref. Section 4). Compare te THD v plot wit te trend for site load current for te corresponding period (ref. Section 4). Calculate te correlation coefficient between THD v and load current (ref. Section 4). Investigate te relationsip between THD v and armonic current levels over te survey period (ref. Section 4.4). Compare te time trends for te site against all oter sites (ref. Section 5). Load Current Victoria June 4: THD vs Load Current Scattergrap 6 5 4 3.5.5.5 THD (%) 4 THDv time series and load current analysis results 4. Commercial sites Te plot of THD v levels against time clearly identified bot daily and weekly cycles of variation. Te plot for one of te commercial sites is sown in Fig.. THDv(%).5.5.5 Victoria June 4 THDv Jun- : Jun- 6:3 Jun-3 3: Jun-4 9:3 Jun-6 : Jun-7 8:3 Jun-8 5: Jun-9 :3 Jun- 4: Jun- :3 Jun-3 7: Jun-4 3:3 Jun-6 6: Jun-7 :3 Jun-8 9: Jun- :3 Jun- 8: Jun- 4:3 Jun-3 : Jun-5 3:3 Jun-6 : Jun-7 6:3 Jun-8 3: Jun-3 5:3 Figure : Montly THD v trend for a commercial On a daily basis, THD v levels peak between te ours of midnigt and 5: a.m., and are fairly constant during tis time at around %. Daily minium THD v levels occur between te ours of 7: a.m. and 5:45 p.m. (at around.5%), and tis coincides wit commercial ours of business and te period of maximum load current. On a weekly scale, weekends (June 5-6, -3, 9- and 6-7) typically experience lower maximum levels and iger minimum levels of THD v. A scatter plot of THD v and load current was used to compare measured THD v levels against corresponding levels of load current for te same period. Tis plot is sown in Fig.. Fig.: Scatter plot of load current against THD v for a commercial Fig. suggests tat a negative correlation exists between load current and THD v. Using te data analysis tools in Microsoft Excel, te correlation coefficient was calculated to be.79 for te mont of June. Very similar results were obtained from te oter commercial site, te only significant difference being te magnitude of te THD v measurements. 4. Industrial Sites Data from eigt industrial sites was analysed in tis study. It is wort noting tat wile tese eigt sites are all considered to ave a predominantly industrial load, some sites are more industrial tan oters (i.e. some sites consist of a mix of industrial, commercial and residential load). Determining wat may be considered representative trends across te eigt sites is furter complicated by te fact tat at tree of te industrial sites, THD v levels are recorded ourly as opposed to te 5-minute recordings at te oter five industrial sites. Despite tese limitations, some trends were still found to be consistent across te eigt industrial sites. Fig.3 sows te THD v trend for wat can be considered a typical THDv(%).8.6.4..8.6.4. Bairds June 4 THDv Jun- : Jun- 3:45 Jun- 3:3 Jun-3 3:5 Jun-4 3: Jun-5 :45 Jun-6 :3 Jun-7 :5 Jun-8 : Jun-9 :45 Jun-3 3: Jun-4 :45 Jun-5 :3 Jun-6 :5 Jun-7 : Jun-8 :45 Jun-9 :3 Jun- :5 Jun- : Jun- :45 Jun-3 :3 Jun-4 :5 Jun-5 : Jun-6 9:45 Jun-7 9:3 Jun-8 9:5 Jun-9 9: Jun-3 8:45 Fig.3: Montly THD v trend for an industrial Te daily and weekly variation in THD v levels typical of industrial sites can be seen in Fig.3. For te site sown, THD v levels peak between 8: p.m. and midnigt. For all industrial sites, THD v levels peak during te nigt-time ours. At some sites tis peak occurs before midnigt, wile at oter sites te peak 87
Presented in AUPEC 7, Pert, Western Australia, 9- December, 7 occurs between midnigt and 5: a.m. Fig.3 also sows a clear weekly trend in THD v levels, wit iger levels being recorded on weekends. Tis is true for most of te industrial sites, and coincides wit periods of lower load current. Te relationsip between measured THD v levels and load current was assessed using a scatter plot. Te scatter plot corresponding to te THD v data from Fig.3 is sown in Fig.4. Load Current 6 4 8 6 4 Bairds June 4 THDv - I Scattergrap..4.6.8..4.6.8 THDv (%) Fig.4: Scatter plot of load current against THD v for an industrial As for te commercial sites, te scatter plot suggests a negative relationsip between THD v levels and load current. For te example industrial site, te correlation coefficient was calculated to be.53. All industrial sites displayed a negative correlation coefficient between THD v levels and load current, wit te coefficients ranging between.34 to.86 Across te eigt industrial sites, te range of peak THD v levels recorded was between.5% to 3%. 4.3 Residential Sites Te Vector power quality survey included data from tree residential sites. Cyclic variation of THD v levels on bot daily and weekly timescales was clearly apparent. Fig.5 sows te variation in THD v levels for one of te residential sites for te survey period. THDv(%) 3.5 3.5.5.5 Takanini June 4 THDv Jun- : Jun- :45 Jun-3 5:3 Jun-4 8:5 Jun-5 : Jun-6 3:45 Jun-7 6:3 Jun-8 :3 Jun- :5 Jun- 3: Jun- 5:45 Jun-3 8:3 Jun-4 :5 Jun-5 4: Jun-6 6:45 Jun-7 9:3 Jun-8 :5 Jun- : Jun- 3:45 Jun- 6:3 Jun-3 9:5 Jun-4 : Jun-5 4:45 Jun-6 7:3 Jun-7 :5 Jun-8 3: Jun-3 :45 Fig.5: Montly THD v trend for a residential For te residential sites, THD v levels typically peak eac day between te ours of 7:3 p.m. and :3 a.m. Witin tis range, most daily THD v peaks occur around :3 p.m. Daily minimum THD v levels typically occur between te ours of 6:3 a.m. and 9: a.m., wic coincides wit te morning peak in load current. During te late afternoon-early evening load current peak, THD v levels also sow a marked decrease. On a weekly time scale, two of te tree sites sowed a clear weekly pattern, wit maximum THD v levels for te week being recorded on eiter Saturday or Sunday. For te tird residential site, te weekly pattern was less clear, but again many of te weekly peak levels occurred on Saturdays or Sundays. Te relationsip between THD v levels and load current is less clear for residential sites tan it is for commercial and industrial sites. For te example residential site sown in Fig.5, a strong negative correlation between THD v and load current was evident for te mont of June, as can be seen from te scatter plot in Fig.6. For tis site, te correlation coefficient was calculated to be.8. Load Current 9 8 7 6 5 4 3 Tkanini June 4: THD vs Load Current Scattergrap.5.5.5 3 3.5 Fig.6: Scatter plot of load current against THD v for a residential For te oter two residential sites, te THD v -load current correlation coefficients were. and.54 (i.e. bot positive). 4.4 Relationsip between THD v and armonic current levels Given tat te presence of armonic voltages on te network is te result of te flow of armonic currents, tis raises te question of ow do te armonic voltage levels relate to measured levels of armonic currents. Te only armonic current data available from eac site are te 5-minute averages of current THD for eac pase. Te usefulness of current THD for comparing values at different sites or at different times is limited, as eac measurement is divided by a different value of fundamental current as te load varies. A more useful value for comparison purposes is te magnitude of te distorting current. Wit data for bot average load current and current THD, te magnitude of te total distorting current can be calculated. Current THD is calculated as: THDv 88
Presented in AUPEC 7, Pert, Western Australia, 9- December, 7 I THDi = I were I I () = total distorting current = fundamental currrent and so te total distorting current can be calculated as armonic current I ave been plotted togeter. Te trend for an example site is sown in Fig.7, clearly illustrating te similarity in variation of THD v and total armonic current. Fig.7: THD v and armonic current I for a sample Bairds June 4 THDv & Harmonic Current I/ V THD mean I = THDi I () For te purpose of tis calculation, only te average rms value of load current (rater tan te fundamental) is available, but given tat I = I + I (3) rms 4.5 4 3.5 3.5.5 THDv & I/ Harm Current I/ it can be sown tat te magnitude of te total distorting current can be calculated by: I Irms THD = + THD i i (4) Te relationsip between THD v and te total distorting current I as been analysed wit te aim of determining weter te source of tis disturbance is armonic load currents flowing downstream of te monitored site, or weter it originates from upstream of te monitored To analyse te relationsip, te correlation coefficient between THD v and total distorting load current I as been calculated for eac Te results are given in Table. Site THDv I Correlation Coefficient Commercial Quay.88 Victoria -.4 Industrial Bairds. 55 Carbine.9 Greenmount.7 McNab. 48 Otara.43 Rockfield.76 Rosebank.9 Wiri.9 Residential Howick.4 Manurewa.75 Takanini.87 Table : Correlation between THD v and d istorting load current I. Te results in Table sow tat all sites except one sow eiter a moderate or strong positive correlation between THD v and total distorting current. Tis suggests tat te voltage distortion levels present at te zone substations are likely due to distorting load currents downstream from te substation. It also indicates tat te variation in armonic current magnitude varies wit time in a similar manner to voltage distortion. To ceck tis, trends in THD v and.5 Jun- : Jun- 9:45 Jun- 5:3 Jun-3 :5 Jun-4 7: Jun-5 :45 Jun-5 :3 Jun-6 8:5 Jun-7 4: Jun-8 9:45 Jun-9 5:3 Jun-3 :45 Jun-3 :3 Jun-4 8:5 Jun-5 4: Jun-6 9:45 Jun-7 5:3 Jun-8 :5 Jun-8 : Jun-9 6:45 Jun- :3 Jun- 8:5 Jun- 4: Jun- 3:45 Jun-3 9:3 Jun-4 5:5 Jun-5 : Jun-6 6:45 Jun-7 :3 Jun-7 :5 Jun-8 8: Jun-9 3:45 Jun-3 9:3 5. Summary of results From comparison of te results from te tree site categories, it is clear tat tere are significant differences in te patterns of THD v variation, and tat tese differences can be attributed to bot predom inant load type, and te pattern of load current variation. Tere are also aspects of THD v variation tat are common across all sites. 5. THD v variation wit time Te first comparison tat can be made is between te times at wic peak THD v levels occur for eac Table sows te typical times of daily peak THD v levels for t e surveyed sites. Te saded areas indicate te range of times of daily peak THD v levels. Te range of peak times is defined as range of times of day during wic te daily maximum value of THD v occurs over te one-mont survey period. 6:-8: 8:-: :-: :-4: 4:-6: 6:-8: 8:-: :-: :-: :-: :-4: 4:-6: Quay Victoria Bairds Carbine Greenmount McNab Otara Rockfield Rosebank Wiri Howick Manurewa Takanini = commerc ial = industrial = residential Table : comparison of THD v peak tim es. 89
Presented in AUPEC 7, Pert, Western Australia, 9- December, 7 Te two zone substations supplying predominantly commercial load bot exibited ve ry similar patterns of variation of T HD v and load current. THD v levels peak between midnigt and 5: a. m., and fall to low levels during business ours wen load current is ig. Te diurnal variation in THD v levels is around %. Overall, THD v levels for te two commercial substations are lower tan for industrial or residential substations. Te average 95% cumulative probability value of THD v across te two sites for te mont of June was.44%. THD v levels at zone substations supplying predominantly industrial loads likewise peak at nigt and are generally iger on weekends, and tis coincides wit periods of reduced load current. Tere is some variation across te eigt substations in te time during te nigt at wic peak THD v levels occur. For most sites te peak occurs between 8: p.m. and midnigt (similar to a residential site) wile at one site te peak occurs between midnigt and 5: a.m. Te average diurnal variation in THD v levels at industrial sites is.6%, similar to tat for commercial sites. Te average 95% cumulative probability value of THD v across te eigt sites was.94%, sligtly iger tan tat for te commercial sites. THD v levels at te zone substations supplying predominantly residential load also tend to peak mostly during nigt-time ours. For two of te residential substations, THD v tends to peak between te ours of 8:3 p.m. and midnigt. Te one residential substation were THD v levels peak after midnigt also as a load current pattern tat suggests tat tis residential site as a significant commercial or industrial load component. Te average diurnal range in THD v levels for te residential substations is.3%, larger tan te range for eiter commercial or industrial sites. Te average 95% cumulative probability value of THD v across te tree residential sites was.58%, significantly iger tan te comparable values for bot commercial and industrial sites. Tis reaffirms te findings of [] tat THD v levels at residential sites are typically iger tan tose experienced at commercial or industrial sites. 5. THD v variation wit Load Current Te second comparison tat can be made between te sites is te relationsip between variation in THD v levels and daily and weekly variations in load current. Te correlation coefficient between THD v and load current for eac site is given in Table 3. Te two commercial sites sow a strong negative correlation between THD v and load current. Likewise, all industrial sites sow a negative correlation between THD v and load current. Te strengt of tis correlation varies between sites, and tis is likely due to variation in te degree to wic te load is purely industrial or combined wit some commercial or residential load. Te most interesting feature of Table 3 is tat two substations display a positive correlation between THD v and load current, and bot of tese substations are supplying predominantly residential load. Site THDv Load Current Correlation Coefficient Commercial Quay -.8 Victoria -.79 Industrial Bairds -.53 Carbine -.38 Greenmount -.86 McNab -.34 Otara -.39 Rockfield -.3 Rosebank -. 78 Wiri -.6 Residential Howick +.54 Manurewa +. Takanini -.8 Table 3: Correlation coeffi cients for THD v against load current. Te daily and weekly cycles of variation in voltage armonic levels ave been observ ed and documented previously in a number of publications, and mention of tis aspect of armonic variation can be found in textbooks on power quality [3], [4]. In [3], a plot of fift armonic voltage against time over a one-week period sows strong similarity wit Fig.5 in tis paper. In [3], it is stated tat te peak voltage armonic levels can be attributed to user beaviour, and in particular te use of televisions. Wile tis may seem a plausible explanation for suc variation on a residential feeder, it does not explain te postmidnigt peak in voltage armonic levels tat occur at commercial and industrial sites. Additionally, at two of te residential sites included in tis study, THD v levels typically peak at between : p.m. and :3 p.m., wereas it would be expected tat te peak time for television watcing would be between te ours of 7: p.m. and : p.m. ( prime time ). In [5] it is concluded tat varying levels of television use ave a surprisingly low impact on armonic voltage levels. It maybe tat relatively iger levels of THD v are experienced during tese periods of low load because of te effective damping effect on armonic distortion tat occurs during periods of iger load current. In [4], it is stated tat iger levels of triplen armonic currents are commonly observed during te early morning ours wen te load is low, and tat tis is due to transformer excitation current. Tis explanation is questionable given tat transformer excitation current is typically muc less tan % of rated full load current. It also does not explain te variation in THD v levels observed in tis survey on a 9
Presented in AUPEC 7, Pert, Western Australia, 9- December, 7 MV network, as triplen armonics do not propagate back into te delta-connected MV distribution network. A possible explanation for te negative correlation between load current and THD v levels observed at most sites is armonic cancellation by te mixing of non-linear single pase and tree pase loads, as described in [6]. It as been found tat mixing single and tree pase loads can result in reduced THD v, te reason being tat te 5 t and 7 t armonic currents of single and tree pase loads are often in counter- pase resulting in at least partial cancellation. It could be suggested tat voltage armonic distortion is not a problem on te network due to te low levels recorded. Typical 95% cumulative probability values across all sites are in te vicinity of % to 3%. Tere is clearly no issue of non-conformance to national or international armonics standards. International standard AS/NZS 6.3.6 [7] specifies a compatibility level of 8% for armonic voltages in MV power systems. Te requirements of te New Zealand Electricity Regulations [8] only relate to measurement of voltage armonic distortion levels at te point of common coupling, and so are not generally applicable to armonic levels on an MV distribution network. Te negative correlation between load current and THD v levels as possible implications for te management and regulation of armonic levels on distribution networks. If te igest levels of distortion occur in te middle of te nigt wen connected load is ligt, tis would suggest tat te impact on most customers (and network equipment) will be minimal. It also raises te question as to weter te application of cumulative probability statistics to specify acceptable armonic levels on distribution networks is appropriate. 6. Conclusions Voltage total armonic distortion levels (THD v ) from a variety of monitored sites on a MV distribution network ave been analysed in terms of variation wit time, predominant load type, and level of load current. It as been found tat for commercial and industrial sites, tere is a consistent negative correlation between THD v levels and load current. For te tree residential sites surveyed, two of te sites ad a positive correlation between THD v and load current. Te remaining site ad a negative correlation between THD v and load current, but tis site also displayed some load caracteristics more typical of a commercial or industrial site rater tan a residential Typical cyclic patterns of variation of THD v ave been identified for eac of te tree main load categories. For all sites, peak THD v levels typically occur at nigt, but weter te peak occurs during te late nigt or in te early morning ours depends on te predominant load type. Weekly patterns of THD v variation were also identified, wit iger levels occurring at weekends. Te findings of tis study cast doubt on previously accepted explanations for iger armonic levels at nigt or early morning. Te effects of television use, transformer excitation current, and te levels of triplen armonics do not satisfactorily explain tis penomena. Furter researc is required into te source(s) of tis armonic disturbance. Te assessing and reporting of armonic levels on MV distribution networks by use of cumulative probability statistics appears to be inappropriate, given tat te igest levels of armonic disturbance occur during periods of low load and wen risk to equipment is lower tan at times of iger load. ACKNOWLEDGEMENTS Te autors would like to acknowledge te support of Vector Ltd (NZ) for allowing access to network power quality data and for supporting tis project. REFERENCES [] Integral Energy Power Quality & Reliability Centre Tecnical Note 3: Harmonic Distortion in te Electric Supply System, Marc, www.elec.uow.edu.au/iepqrc/files/tecnote3.pdf [] Nicolson G, Gosbell V, Parsotam A, Te Influence of Site Pysical Caracteristics on Power Quality Performance, Proc. AUPEC 6, December 6, Melbourne Australia. [3] Sclabbac J., Blume D., Stepanblome T., Voltage Quality in Electrical Power Systems, IEE, London. [4] Dugan R., McGranagan M., Beaty H., Electrical Power Systems Quality, McGraw- Hill, New York, ISBN -7-83-8 [5] Browne N., Perera S., Ribeiro P.F., Harmonic Levels and Television Events presented at IEEE PES General Meeting, 4-8 June 7, Tampa Florida, USA [6] Hansen S., Nielsen P., Blaabjerg F., Harmonic Cancellation by Mixing Nonlinear Single-Pase and Tree-Pase Loads, IEEE Transactions on Industry Applications, Jan pg.5-59 [7] AS/NZS 6.3.6: Limits Assessment of emission limits for distorting loads in MV and HV power systems. [8] New Zealand Electricity Regulations 997, www.ess.govt.nz/rules/pdf/electricity_regulation s_997.pdf 9