The Effect of PV on Transformer Ageing: University of Queensland s Experience

Transcription

1 Australasian Universities Power Engineering Conference, AUPEC 214, Perth, WA, Australia, 28 September 1 October The Effect of PV on Transformer Ageing: University of Queensland s Experience D. Martin, S. Goodwin, O. Krause, T. Saha Power and Energy Systems ITEE University of Queensland Brisbane, Australia Abstract Rooftop PV has an advantage that it can reduce the load supplied by nearby distribution transformers. This in theory should make the transformer run cooler, and therefore extend the remaining life of its paper insulation. However, the relative impact of harmonics on transformer life is sometimes neglected. In this article we present the findings of an analysis on the new rooftop PV installation at University of Queensland, in Brisbane. The load profile of three transformers was investigated, with IEC thermal models applied to estimate hot-spot temperature. The relative effect of PV on the weighted hot-spot temperature was investigated, and then a determination made on the likely loss of life. The harmonics injected by the PV was then explored. Keywords paper insulation; photovoltaic systems; power transformers; power system harmonics; remaining life assessment I. NOMENCLATURE Pre-exponential ageing coefficient (h -1 ) Difference operator Degree of polymerization Activation Energy (J mol -1 ) h Harmonic order RMS fundamental load current (A) RMS current at harmonic h (A) Rated current (A) K Load factor Thermal model constant Thermal model constant Thermal model constant Winding eddy-current loss (W) Winding eddy-current losses at the measured current and the power frequency (W) Ratio of load losses at rated current to no-load losses Ideal gas constant (J K -1 mol -1 ) Temperature ( C) Oil exponent Winding exponent Ambient temperature ( C) Hot-spot temperature ( C) Hot-spot temperature for a given interval of time ( C) Weighted hot-spot temperature ( C) Top-oil temperature at the load considered ( C) Average oil time constant (min) Winding time constant (min) Hot-spot to top-oil gradient at the load considered (K) Hot spot to top oil gradient at rated current (K) Top oil temperature rise in steady state at rated losses (K) II. INTRODUCTION The University of Queensland (UQ) has heavily invested in photovoltaic (PV) electricity generation, installing a large number of panels on building rooftops. This system became operational in June 211. The PV lowers the power demand of these buildings, and of the adjacent areas. One advantage is that lowering the demand of these buildings will result in the supplying transformers being less loaded, which should therefore reduce their internal temperature making them last longer. This effect was modeled for an Australian city by Pezeshki and Wolfs [1], who considered the extension of transformer paper life to be predominantly due to offsetting the current supplied by the transformer. They used the IEEE loading guide [2] to estimate life. One problem with the IEEE loading guide is that it does not take account the oxygen and water levels within the transformer, which accelerate the rate of ageing, an effect studied in detail [3-8]. PV array inverters are known to inject harmonics [9-1], which are only simplistically taken into account by thermal models. Harmonics cause higher eddy current losses in winding conductors and structural parts, consequently resulting in higher operating temperatures [11]. The IEC standard for transformers, IEC 676/1, regards that under normal service conditions, the total harmonic current does not exceed 5 % of rated current [12]. If it does, the temperature rise will increase and may exceed rated rise. In this article the effect of the PV, and harmonics, on transformer life was analyzed using an updated ageing model. III. THE UNIVERSITY DISTRIBUTION NETWORK As of 214 UQ had the largest integrated PV installation in Australia. 1.2 MW has been installed across four buildings covering 8,2 m 2 of roof space. During peak usage 5 6 % of the campus s electricity requirement is met from PV. The university receives an 11 kv supply from the local utility, which is distributed around the medium voltage network of the campus. Forty substations are used to step the voltage down to normal mains level supply. An extensive network of online digital power meters is used to monitor the electricity infrastructure, which can be interrogated over the university intranet.

2 Australasian Universities Power Engineering Conference, AUPEC 214, Perth, WA, Australia, 28 September 1 October Two multi-story car parks each have 235 m 2 of rooftop PV installed, each giving a nominal DC output of 339 kwp (Fig 1). Each car park uses 26 inverters (24 x 12.5 kw and 2 x 5 kw) to export power. Given that the load required by these car parks is relatively low, only for lighting, most of the generated power is exported to the adjacent buildings. In this substation two 11 kv/415 V 75 kva transformers are used, the nameplate rating of the one used in this study is given in Table 1. Under normal operation one transformer is connected to one car park PV array. Power quality meters monitor the harmonics on both the low voltage side of the transformer, and at the output of the PV inverter. The power quality meter connected to the transformer under study gave a metric of the harmonic spectrum in terms of K-factor, where the K-factor is the sum of the squares of the per unit harmonic current times the harmonic number squared (1) [11], [13]. Fig. 1. Solar panels on identical car parks. (1) IV. TEMPERATURE DETERMINATION The IEC thermal model [14] was used to estimate both top oil and hot-spot temperature, with and without contribution from the PV. Thermal parameters for these types of transformer were used as suggested in the IEC 676 model. For the without-pv case it was assumed that the transformer load could be calculated by summing the measured load with that supplied by the PV. The effect of harmonics on transformer loss of life was investigated by Elmoudi [15]. They found for their test transformer that a total harmonic distortion (THD) of 1 % raised the internal temperatures by a few degrees, whereas, a THD of 22% can raise the temperature levels by more than 1 C. Given that the life of transformer paper insulation is halved for every 6 C, harmonics can considerably shorten life. The current IEC thermal model does not take into account harmonics. Equations (2-4), from the IEC standard [14], are used to calculate initial conditions: (2) (3) 1 (4) V. PAPER AGEING MODELING TAKING INTO ACCOUNT EFFECT OF HARMONICS BY DE-RATING TRANSFORMER A year of loading data was downloaded for analysis, measured at half-hour intervals. The general approach was to use the IEC thermal model to calculate hot-spot temperature for a transformer, with and without contribution from the PV, and compare the average rate of ageing. TABLE I. TRANSFORMER NAMEPLATE RATING AND THERMAL CHARACTERISTICS Rating Cooling mode and preservation system 11 kv/415 V, 75 kva Secondary current: 17 A ONAN Free breathing + desiccator The change in top oil temperature on a time step is given by (5). (5) The temperature at a given time is therefore: (6) The hot-spot temperature is calculated using: (7) / 1 (8) (9) (1) (11) (12) The models proposed by the IEC and IEEE to determine the life remaining of paper are simplistic because they do not take into consideration the effects of water and air; whereas the chemical kinetic model used in this study does, detailed below. A. Chemical kinetic model This model is based on work published by Ekenstam [5], Emsley [6], Lundgaard [7] and Lelekakis [3,4]. The time required for the degree of polymerization (DP) of paper to fall to another level is calculated from the chemical conditions within the transformer, (13).

3 Australasian Universities Power Engineering Conference, AUPEC 214, Perth, WA, Australia, 28 September 1 October (13) Subscripts and t refer to the DP value at the start and at time t. The A-value is dependent on the availability of water and oxygen. A free breathing transformer will have a high oxygen content. Equations (14) to (16) are used to estimate the A- value, where w is the paper water content. The weighted hot-spot temperature was calculated for the transformer, with and without PV, to determine how the reduction in transformer load affects its life. The weighted temperature fell from 59 C to 5 C. For O 2 < 7, ppm: (14) For 7, < O 2 < 14, ppm: (15) Fig. 2. Load levels for transformer, with and without offsetting by PV. For 14, < O 2 < 25, ppm: (16) B. Measured Current Levels and load The current profiles, averaged both for all three phases and over 3 minutes, were recorded. Fig 2 shows the load profile of the transformer, the no PV trace shows the load if no current was being provided from the PV array. Fig 3 shows 5 the level of current supplied by both PV and transformer over a 5 day period. It can be seen that proportionally, compared to the overall electricity requirement, the array generates a significant proportion of the total power requirement. C. Thermal Modeling IEC thermal modelling equations (2 12) were used for comparing the respective hot-spot temperature, with or without PV. Given that the time between load measurements should be half that of the winding time constant ( 3 minutes), although the measurements were actually taken every 3 minutes, values for load were interpolated linearly for every 1.5 minutes. The ambient temperature data was sourced from the local university weather station, operated by the School of Geography Planning and Environmental Management, which records data minutely (given in Fig 4). The load and ambient weather measurements were used as inputs into the IEC thermal model to determine the hot-spot temperature profile, shown Fig 5. D. Life of Paper Insulation Equation (17) was used to determine the weighted hot-spot, which is a constant temperature that produces the same ageing effect over the same ageing time as the hotspot temperature profile [8], n denotes number of measurements. 273 (17) Fig. 3. Current level for car park 2, showing the current supplied by the transformer, the current supplied by PV and the total current. Fig. 4. Ambient temperature as measured by the campus weather station. Fig 6 shows the change in life of paper. Since the water content is not currently known, a distribution is shown. Since the transformer is free breathing is likely to have wet insulation. The unit has a desiccant breather to help prevent condensation within the tank. But, this will not prevent oxygen reacting with the oil producing acids and water. The current offsetting can be seen to significantly extend the remaining life of the transformer.

4 Australasian Universities Power Engineering Conference, AUPEC 214, Perth, WA, Australia, 28 September 1 October kwp plant in Northern Greece [16]. The group showed that the magnitude of each harmonic was somewhat scattered as a function of the fundamental current. Investigating the first 11 harmonics may also be simplistic, because Papathanassiou [17] measured up to the 5 th harmonic for an inverter (used by a wind turbine), which would not have been captured by Papaioannou. Although the magnitude of the harmonic generally falls at the higher harmonic orders, the impact is related to the square of the frequency (19), which could become significant at very high frequencies. Fig. 5. Calculated hot-spot temperature for transformer. Life expectancy (years) Change in temp Water content of paper.5% 1.% 1.5% 2.% 3.% B. Harmonic voltages measured at UQ The PV array was divided into three sub-arrays, each being individually metered. Meters provided half-hourly readings for voltage and current THD. These meters were however unable to measure the individual harmonics. The THD of voltage measured at the inverter of the PV array is shown in Fig 7. This has remained fairly low throughout the year. Although there are peaks in the morning and just before midnight, these peaks do not appear to be consistently at the same time over the year. For instance, the peak around midnight is only observed in the first days of the year Temperature ( C) Fig. 6. Change in life expectancy of paper of transformer, as a function of paper water content. VI. INVESTIGATING THE EFFECT OF HARMONICS The IEEE standard on non-sinusoidal loads [11] recommends measuring the harmonic currents at full load and de-rating the transformer accordingly. This is problematic if the spectrum of the harmonics change, then so will the degree of effect. The new rating is then used with the IEC thermal model [12, 14]. One measure of harmonics is the total harmonic distortion (THD), (18). Total harmonic distortion of voltage (%) Time (hours) Days.2 % A. General Effect of Harmonics The general effect of a harmonic on the power dissipated by the winding conductor is given in (19). It is important to note that the relative effect of each harmonic number is dependent on the square of its frequency. (18) watts (19) One problem with this method is that the harmonic spectrum from the inverters is not constant. Papaioannou investigated the first 11 odd harmonic currents from a small Fig. 7. THD of current of PV sub-section, as a function of time of day (x-axis) and day of year (z-axis). The THD is highest around sunrise and sunset, however, the fundamental current at these times is low (Fig 8). C. Harmonic currents measured at UQ The THD from the PV array increases around sunrise and sunset (Fig 8), when the current output from the PV is low (Fig 9). Consequently, although the THD may be high during certain times the level of current involved is small. Thus, the extra heating effect within the transformer could be minimal. Fig 1 show the relationship between current output and THD. It can be seen that THD is only high within a small band of current level, up to about 2 amps. An implication is that if this PV array were used in a location where the solar irradiance frequently drops to levels sufficient to generate this low current, there could be more harmonics.

5 Australasian Universities Power Engineering Conference, AUPEC 214, Perth, WA, Australia, 28 September 1 October Total harmonic distortion of current (%) Time (hours) 2 Days The sum of the RMS harmonic currents was calculated from the fundamental current and THD by re-arranging (18), and is shown in Fig 11. The THD is not only higher because of the very low fundamental current, but also because the harmonic currents themselves are also 3 4 times higher in this region than normal. Fig. 8. THD of current of PV sub-section, as a function of time of day (x-axis) and day of year (z-axis). The THD is highest around sunrise and sunset, however, the fundamental current at these times is low (Fig 8). THD is averaged over 3 minute blocks. Current (A) Time (hours) Fig. 9. Current output of PV array, hour of day is on x-axis, day of year along z-axis. Current is averaged over 3 minute blocks Days Fig. 11. The relationship between the magnitude of the fundamental current, and the magnitude of the sum of harmonic currents. These are highest when the fundamental is between 5 and 7 amps. Measurements are averaged over 3 minute blocks. In order to investigate how both the THD and K-factor varied over the sunrise hours, they were recorded for one morning on a 5 minute resolution, over a 3 hour period, on a cloudy day. The solar irradiance and currents are shown in Fig 12. The solar irradiance was measured by a weather station operating at UQ, not of the PV itself, possibly accounting for the slight shifting in time between the readings. Both the THD and K-factor of the current from the PV array were very high while its current output was low, shown in Fig 13. However, in proportion to the overall current carrying capacity of the transformer, the harmonics from the PV have had little effect on its THD and K-factor. Average current transforme er Average current PV Solar irradiance Current (A) Irradiance (W/m 2 ) Fig. 1. The relationship between the current generated by the PV and the total harmonic distortion. Measurements are averaged over 3 minute blocks. 5:45 6:57 8:9 9:21 Fig. 12. Comparing the load of the transformer, output of the PV and solar irradiance at sunrise (5:55 am). Measurements averaged for the 3 phases.

6 Australasian Universities Power Engineering Conference, AUPEC 214, Perth, WA, Australia, 28 September 1 October THD (%) Highest transformer THD% Highest K-factor transformer Highest PV THD% Highest K-factor PV 5:45 6:57 8:9 9:21 Fig. 13. Investigating how the THD and K-factor for the PV inverter changes at sunrise (5:55 am), and compares with values for the transformer. Highest denotes that the measurement was the highest of the three phases. Although the effect on this transformer appears insignificant, possible effects of high PV penetration on smaller residential transformers should be considered. For instance, most utilities permit a PV installation of between 5 to 1 kw without a technical assessment being required [18]. A utility may base the number of customers to be connected to a transformer on an After Diversity Maximum Demand (ADMD), which can be only 4.5 kw per customer [19]. Given that the installed PV capacity of a street can be larger than the capacity of the transformer, this can magnify the effect of harmonic currents. For instance, the K-factor of the transformer shown in Fig 13 appears unaffected by the PV because the harmonic currents involved are relatively minor. However, if the transformer had a lower rating then proportionally the harmonic currents would be more significant. This UQ PV array, 339 kwp, is very large compared to a residential rooftop PV. However, since 26 inverters are used this installed capacity could be representative of a street with a significant penetration of PV. The effect of harmonics at low current levels should be explored further to determine whether they have a significant impact on the overall life of small distribution-transformer insulation, especially during low light levels. VII. CONCLUSIONS The calculations for paper life indicated longer life, for this transformer, approximately doubling the remaining life. The THD injected by the invertors of the PV array was much higher, up to over 8 %, during low levels of solar irradiance, < 1 W/m 2. In Brisbane this occurred at sunrise and sunset. A PV array located in an area with less solar irradiance may therefore produce comparatively more harmonics. Analysing the harmonic and fundamental currents showed that high THD was not only due to low output current, the sum of the harmonic currents were also 2 3 times higher. The effect of harmonics on the power loss within a transformer is dependent on the square of its frequency in K factor addition to its current. The meter used could not determine the spectrum of the harmonics, and so only K-factors could be used. Given that the K-factor changed during sunrise, it was not possible to determine whether it was only the magnitude of the harmonic currents which were changing, or if their distribution was also affected. Some suggestions to reduce or compensate a high THD include: requiring an inverter manufacturer to specify THD at low power in addition to at rated current, altering standards to include a limit for THD at low current, and, including a filter designed to remove harmonics during low current output. VIII. REFERENCES [1] H. Pezeshki and P. Wolfs, Impact of High PV Penetration on Distribution Transformer Life Time, IEEE Int. Symp. On Electrical Insulation, pp , Canada, 26. [2] IEEE Guide for Loading Mineral-Oil-Immersed Transformers, IEEE Std C (R24), USA, 24. [3] N. Lelekakis, D. Martin, and J. Wijaya, Ageing rate of paper insulation used in power transformers. Part 1: oil/paper system with low oxygen concentration, IEEE Trans. Dielectr. Electr. Insul., Vol. 19, pp , 212. [4] N. Lelekakis, D. Martin, and J. Wijaya, Ageing rate of paper insulation used in power transformers. Part 2: oil/paper system with medium and high oxygen concentration, IEEE Trans. Dielectr. Electr. Insul., Vol. 19, pp , 212. [5] A. Ekenstam, The behaviour of cellulose in mineral acid solutions: kinetic study of the decomposition of cellulose in acid solutions, Berichte der deutschen chemischen Gesellschaft, Vol. 69, pp , [6] A. M. Emsley and G. C. Stevens, Kinetics and mechanisms of the lowtemperature degradation of cellulose, Cellulose, Vol. 1, pp , [7] L. E. Lundgaard, W. Hansen, D. Linhjell, and T. J. Painter, Aging of oil-impregnated paper in power transformers, IEEE Trans. Power Delivery, Vol. 19, pp , 24. [8] N. Lelekakis, W. Guo, D. Martin, J. Wijaya, and D. Susa, A field study of ageing in paper-oil insulation systems, IEEE Electrical Insulation Magazine, Vol. 28, Iss. 1, pp , 212. [9] A. A. Latheef, V. J. Gosbell, and V. Smith, Harmonic Impact of Residential Type Photovoltaic Inverters on 11 kv Distribution System, Australasian Universities Power Engineering Conference, Australia, 26. [1] Z. Xiangyang and L. Shiyang, A Research of Harmonics for Multiple PV Inverters in Grid-Connected, Power and Energy Engineering Conference, pp. 1 4, China, 212. [11] IEEE Std C (Revision of IEEE Std C ) IEEE Recommended Practice for Establishing Transformer Capability When Supplying Non-sinusoidal Load Currents, [12] IEC 676-1, Power transformers Part 1: General, Switzerland, 211. [13] G. W. Massey, Estimation Method for Power System Harmonic Effects on Power Distribution Transformers, [14] IEC 676-7, Power transformers - Part 7: Loading guide for oilimmersed power transformers, Switzerland, 25. [15] A. Elmoudi, M. Lehtonen, and H. Nordman, Effect of harmonics on transformers loss of life, IEEE Conference on Electrical Insulation, pp , Canada, 26. [16] I. Papaioannou, A. Bouhouras, A Marinopoulos, and M. Alexiadis, Harmonic Impact of Small Photovoltaic Systems Connected to the LV Distribution Network, IEEE Conference on European Electricity Market, Portugal, 28. [17] S. Papathanassiou and M. Papadopoulos, Harmonic Analysis in a Power System with Wind Generation, IEEE Transactions on Power Delivery, Vol. 21, Iss. 4, pp , 26. [18] B. Noone, PV Integration on Australian Distribution Networks, The Australian PV Association, UNSW, Australia, 213. [19] Supply & Planning Manual, Energex, Australia, 213.