- Email: [email protected]
On the impact of single-phase plug-in electric vehicles charging and rooftop solar photovoltaic on distribution transformer aging
Electric Power Systems Research 148 (2017) 202–209
Contents lists available at ScienceDirect
Electric Power Systems Research journal homepage: www.elsevier.com/locate/epsr
On the impact of single-phase plug-in electric vehicles charging and rooftop solar photovoltaic on distribution transformer aging M.K. Gray, W.G. Morsi ∗ Faculty of Engineering and Applied Sciences, UOIT, Oshawa, ON, Canada
a r t i c l e
i n f o
Article history: Received 13 December 2016 Received in revised form 20 February 2017 Accepted 23 March 2017 Keywords: Monte Carlo methods Plug-in electric vehicles Power quality Rooftop solar photovoltaics Transformer aging
a b s t r a c t This study investigates the impact of single-phase plug-in electric vehicles charging on increasing the rate at which center-tapped distribution transformers experience aging. Distribution transformer aging is investigated considering varying rooftop solar photovoltaic generation penetration rates. Monte Carlo methods are used to probabilistically estimate the transformer’s loss of life considering the effect of timeof-use (TOU) pricing. The results of applying the proposed method have revealed that plug-in battery electric vehicle charging impact on both transformer aging and neutral current is largest in the case that vehicles charge based on time-of-use pricing methods. Further application has shown that while rooftop solar photovoltaic generation reduces transformer aging, no significant reduction in neutral current is observed. © 2017 Elsevier B.V. All rights reserved.
1. Introduction 1.1. Background In 2009, the Government of Ontario launched the micro feedin-tariff (microFIT) program [1], allowing homeowners to generate power from rooftop solar photovoltaic systems (PV) up to 10 kW, and get paid for the energy they produce over a 20-year term. Through adding such financial incentives for residential homeowners, the MicroFIT program aims to increase renewable energy generation in the province of Ontario. The Government of Ontario also launched the Ontario’s Electric Vehicles (OEV) incentive program and Green License Plates [2] in 2013, through which Ontario residents receive up to $8500 in rebates for the purchase or lease of new plug-in battery electric vehicles (PBEVs). 1.2. Problem statement While both microFIT and OEV incentives look to increase the penetration of Green technologies to reduce greenhouse gas (GHG) emissions in Ontario, the authors’ previous work [3] has shown that charging the battery of electric vehicles using Level 2 (240 V) charging may significantly increase distribution transformer loading for several hours, resulting in reduced transformer lifetimes. Studies [4,5] have shown increased rooftop solar PV penetration may
∗ Corresponding author at: University of Ontario Institute of Technology, Oshawa, ON L1H 7K4, Canada. Fax: +1 905 721 3370. E-mail address: [email protected] (W.G. Morsi). http://dx.doi.org/10.1016/j.epsr.2017.03.022 0378-7796/© 2017 Elsevier B.V. All rights reserved.
reduce loading on distribution transformers feeding the secondary system, and hence extending their lifetime; however these studies have not considered the split-phase nature of North American secondary systems, and therefore have not investigated the effects of rooftop solar generation considering Level 1 (120 V) plug-in electric vehicle charging. The unbalanced loading caused by electric vehicle charging at Level 1 (120 V) in residential homes has been investigated in Refs. [3,6]. The results of these studies have outlined increased neutral current and reduced distribution transformer’s lifetime as significant factors in considering split-phase electric vehicle charging impact. As the transformers lifetime is given as the split-phase winding which degrades the fastest, unbalanced loading due to PBEV charging reduces transformer lifetime faster than if the loading was balanced. In this respect, there is a need to quantify the effect of single-phase PBEV charging demand on the aging of distribution transformer’s life whilst considering rooftop solar PV generation.
1.3. Work to date While many works investigate the impact of electric vehicle charging on distribution transformer lifetime, a large number of studies [7–15] do not consider the effects of Level 1 electric vehicle charging; which charges using less power than Level 2 charging, for a longer duration of time. Studies considering Level 1 electric vehicle charging either do not consider the North American splitphase distribution transformer [16,17], or do not consider the effects of rooftop PV generation [3,6,18,19]. In this regard, no
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
203
Table 1 Transformer loss of life parameters.
Fig. 1. Center-tapped transformer connections to secondary system.
studies have investigated the mitigation potential of rooftop PV generation on reducing the impact caused by Level 1 electric vehicle charging. 1.4. Contribution
Parameter
Value
A H,R TO,R TO W m n R
30 ◦ C 27 ◦ C 53 ◦ C 6.86 h 0.08 h 0.8 0.8 4.87
Where A is the ambient temperature, H,R is the hottest-spot conductor rise above the top-oil temperature under rated load, TO,R is the top-oil rise above the ambient temperature under rated load, TO and W represent the top-oil and winding thermal time constants respectively, n and m are empirical top-oil rise and winding constants, and R is the ratio of load loss at the rated load to the no-load losses.
Table 2 Secondary distribution system extension nodes.
This paper presents a probabilistic approach using Monte Carlo to estimate the loss of life and neutral current impact on distribution transformers due to PBEV charging in an active distribution system. The power generation from rooftop solar photovoltaic is probabilistically estimated after modeling the secondary circuit components (e.g. transformer, service lines and service drops) feeding residential homes at which PBEVs charging is taking place. The approach presented in this study investigates the effect of time of use prices applied to electric vehicle charging as quantified in terms of distribution transformer aging and neutral current. Furthermore, the results of this study consider the impact reduction potential of rooftop solar PV penetration to reduce the impact of PBEV charging.
Node
Original load rating
Transformer rating
822 Phase A 846 Phase B 862 Phase B
50.69 kVA 17.00 kVA 20.87 kVA
50 kVA 25 kVA 25 kVA
Number of residential homes 10 6 6
2. Distribution transformer impact metrics 2.1. Distribution transformer neutral current Fig. 1 shows the equivalent circuit diagram of the center-tapped distribution transformer windings used to feed secondary circuits. When the currents drawn by split-phases A and B are not equal the presence of a neutral current IN may be observed. Kirchhoff’s Current Law (KCL) expresses this resultant transformer neutral current as the vector sum of the transformer split-phase currents in (1). In = Ia − Ib
(1)
Fig. 2. Ten house secondary system.
Transformer lifetime aging depends on the hottest-spot temperature, denoting the point of insulation on the transformer which deteriorates the fastest. Table 1 lists the thermal parameters needed for LoL calculations on a distribution transformer with 50 kVA nameplate rating [3].
where In , Ia , and Ib represent vectors of the neutral, split-phase A, and split-phase B currents measured at the transformer.
3. Probabilistic simulation methodology
2.2. Distribution transformer aging
3.1. Modified IEEE 34 Bus Test Distribution System
The IEEE Standard C57.91-2011 [20] details the standardized methodology used to calculate the percentage loss-of-life (LoL) of a given transformer based on the transformers hot-spot temperature.
Typically, distribution systems in North America consist of a primary system which extends secondary circuits used to feed residential customers. The primary system modeled in this work is taken as the IEEE 34 Bus Test Distribution System [21] using the exact lumped load model detailed in Ref. [22]. In order to accommodate secondary circuits, the load at each node listed in Table 2 (with apostrophes denoting intermediate node extension based on the lumped load model) was extended with a corresponding secondary circuit as follows. Single phase loads listed in Table 2 were removed and replaced with corresponding center-tapped transformers as rated in Table 2. Each transformer was further extended with a secondary circuit used to feed homes, with circuit model based on the archetype design in Ref. [23]. An example layout of the 10-house secondary distribution system (SDS) fed from a 50 kVA distribution transformer is depicted in Fig. 2 with residential house nodes labeled nH1 to nH10 . As detailed in Ref. [23], secondary lines (SL) and service drops (SD) are assumed 4/0 AA and 1/0 AA triplex cables respectively.
LoL(%) =
FEQA × t × 100 Normal Insulation Life
(2)
where the normal insulation life of a transformer is typically 180,000 h [20], t is the total number of hours in a day, and FEQA is the average aging factor. FEQA =
N
FAA,n tn
N /
n=1
tn
(3)
n=1
Given time step size tn , the aging acceleration factor FAA,n determines the increase in transformer lifetime degradation based on the winding hottest-spot temperature H . FAA = e
15000 383
−
15000 H +273
(4)
204
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
Table 3 Vehicle and charger parameters.
Table 4 Temperature factor table.
Parameter
Value
T (◦ C)
FT (T)
εv Cv SOCmin c
0.24 kWh/mile 24 kWh 5% 0.90
0 25 75 100
1.2 1.0 0.8 0.6
3.2. Residential load profiles The IEEE Reliability Test System (RTS) [24] details a representative load profile given as a percent of the peak load at an hourly resolution over a year. Through sampling the day of the year with uniform random probability, the daily load profile for each residential home in the secondary is obtained from the IEEE RTS, assuming a peak load of 6.64 kVA with 0.9 lagging power factor [25]. Residential loads are assumed constant power loads, with hourly power expressed as: Sn (h) = SnP · Pn (h)
(5) SP
where Sn (h) is the apparent power of house n at hour h, n is the peak apparent power of house n, and Pn (h) is the percent of peak load experienced by house n at hour h for the day sampled from the IEEE RTS.
depends on the rating of the PV array PPV,Rated in kW, the solar irradiance Irad in kW/m2 and the corresponding temperature factor FT (T) for temperature T linearly interpolated from Table 4 (8). PPV,Array = PPV,Rated × Irad × FT (T )
(8)
The power output from rooftop solar photovoltaics (PV’s) to the alternating current (AC) grid PPV,Out is calculated by considering the inverters efficiency Inverter applied to the direct current (DC) power generated from the array. Inverter efficiency is calculated based on the per unit output of the PV array in Table 5 [29]. Hourly irradiance and temperature seen by each rooftop solar PV are taken from the Canadian Weather Energy and Engineering Datasets (CWEEDS) 2005 data in Toronto, Ontario [30] based on the uniform randomly sampled day of the year. The number of PVs in the system is as described in Section 3.5. 3.5. Monte Carlo algorithm
3.3. Plug-in battery electric vehicles charging demand Plug-in battery electric vehicle charging demand is estimated in this study using the model developed in Ref. [26], which calculates total electric vehicle charging energy considering the state-ofcharge (SOC) of the electric vehicle’s battery and the efficiency of the electric vehicle charger. The SOC of the electric vehicle depends on the distance the electric vehicle has driven as well as the electric vehicle’s mileage in terms of energy consumption. Vehicle driving distance is sampled from the National Household Travel Survey (NHTS) 2009 [27] which reports annual mileage (ANNMILES) of surveyed vehicles. The daily distances traveled were obtained by normalizing the annual mileage through division by 365 representing the days in a year. The total energy required by each vehicle v to fully charge is then calculated using (6) [3].
Ev =
max dv × εv , Ev,min
C
(6)
With Ev the total energy required to fully charge vehicle v, dv the sampled driving distance (miles), εv the energy consumption (kWh/mile), C the charger efficiency (per unit), and Ev, min the largest amount of energy required by the vehicle when the battery is at minimum SOC calculable in (7) [3]. Ev,min = Cv × SOCmin
A Monte Carlo simulation is used in this work to address the uncertainties associated with PBEV charging (e.g., daily driving distances, charging time, etc.) and rooftop solar PV generation (e.g., daily temperature and solar irradiance). Through performing a number of independent random experiments (Monte Carlo trials) the impact metrics are taken as the expected value of the statistical results. The simulation in this work considers two PBEV penetrations: no PBEV in the system and one PBEV at each house. Driving distance for each PBEV in the system is sampled using NHTS outlined in Section 3.3, which begin charging based on the Vehicle Charge Start Time (VCST). Simulations are performed for four separate VCST schemes: home arrival time (HAT), time of use 7 p.m. (TOU-7 P.M.), time of use midnight (TOU-midnight), and smart charging (SC) as described in the following subsections.
(7)
where Cv is the battery capacity of the vehicle and SOCmin is the minimum state of charge of the vehicle in per unit. All vehicles in this study are assumed Nissan Leaf, with vehicle and charger parameters outlined in Table 3. Electric vehicle charging on the distribution system further assumes a constant power charging model at unity power factor, using Level 1120 V charging rated 1.44 kW based on SAE J1772 electric vehicle charging standards [28]. The number of vehicles in the system is varied as outlined in Section 3.5. 3.4. Rooftop solar PV power generation model Residential rooftop solar PV installations consist of a solar array, which generates DC power, the power of which is then passed through an inverter to connect to the AC distribution system. The DC power generated from residential rooftop PV arrays PPV,Array
3.5.1. Home arrival time (HAT) Home arrival time charging assumes each PBEV owner will begin to charge their vehicle based on the time at which the PBEV owner returns home from work. The probability distribution for HAT charging is derived from the National Household Travel Survey (NHTS) [27] as follows. Following the NHTS Codebook [31], the subset of vehicles flagged as returning home from work are extracted from the original dataset, and a probability distribution is created from the vehicle trip “ENDTIME” to determine the hour at which a vehicle begins charging. The resultant probability density function (PDF) for vehicles home arrival times is depicted in Fig. 3, with the most common hour of arrival at 6 p.m. (i.e., hour 18 of the day). In the case that PBEV charging begins based on HAT, the home arrival time PDF is sampled for each PBEV in the system to determine when the PBEV begins charging. 3.5.2. Time of use (TOU) 7 p.m. Time of use charging reflects the notion that consumers will strive to minimize the cost of vehicle charging, by taking advantage of the reduced electricity rates according to time of use (TOU) in Ontario, Canada [32]. In this respect, the start of off-peak pricing begins at 7 p.m., in which all vehicles are assumed to begin charging simultaneously.
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
3.5.3. Time of use (TOU) midnight While TOU off-peak pricing begins at 7 p.m., this time coincides with peak loading of residential homes. In order for electric vehicle owners to maintain low electricity prices to charge vehicles and reduce the risk of overloading the transformer; TOU-midnight assumes a hypothetical scenario in which off-peak prices for PBEV charging begin at midnight. In this respect, all vehicles begin charging at midnight, a time at which residential loading is off-peak. 3.5.4. Smart charging The smart charging scheme aims to minimize the expectedly higher impact of PBEV charging on the distribution system simultaneously by splitting vehicle charging equally into two groups [15]. In this respect, the first group of PBEV on the system begin charging at 11 p.m., whereas the second group of PBEV begin to charge starting at 3 a.m. Through splitting vehicle charging into two groups, the SC approach aims to allow customers to charge vehicles during the off-peak TOU pricing, with reduced impact on the distribution system.
Rooftop solar PV arrays are assigned equally to all households from 0 kW/house to 10 kW/house based on MicroFIT limits [33], with a step size of 5 kW. The percentage PV penetration varies from 0% (i.e., 0 kW/House) to 100% (10 kW/House). Hourly temperature and irradiance data are sampled for each Monte Carlo trial based on CWEEDS data outlined in Section 3.4. House load profiles are sampled once for each secondary system of a Monte Carlo trial using IEEE RTS as outlined in Section 3.2. The Monte Carlo simulation is further detailed in Algorithm 1. The number of Monte Carlo trials required by the simulation to converge has been found to be 6000 following visual inspection based on mean transformer LOL, through which additional Monte Carlo trials resulted in negligible change to mean annual transformer LOL.
Algorithm 1.
Monte Carlo simulation procedure
1: Start Input parameters: Home arrival time and daily driving distance CDFs Temperature and irradiance CDF’s Nissan Leaf Parameters (εv,Cv,SOCmin,ηC) Number of Monte Carlo Trials (NMC) 2: For trial = 1:1:NMC 3:
For PBEV/House = 0:1
4:
For each vehicle v
5:
Assign vehicle to house
6:
Determine VCSTv (HAT, TOU-7pm, or TOU-midnight)
7:
Sample dv from daily driving distance
8: 9:
Calculate Ev using (6) End for
10: Assign kW rating to PV at each house 11: Sample daily temperature and irradiance at hourly resolution using CWEEDS 2005 data in Toronto, Ontario 12:
For t=1:24
13:
Perform backward/forward sweep
14:
Record neutral current and center-tapped distribution transformer annual loss of life
15: End for 16: End for
17: End for 18: Evaluate Monte Carlo data 19: End
205
206
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
Fig. 4. Split-phase component model.
Forward/backward sweep is repeated for each hour of the day as specified previously in Algorithm 1. p
p
Vi (pu) = Fig. 3. Probability density function for sampling the hour at which a PBEV arrives home from work (i.e. begins to charge). Table 5 Inverter efficiency table. FInverter (PPV,pu )
0.1 0.2 0.4 1.0
0.86 0.90 0.93 0.97
5.1. Transformer loss-of-life
4.1. Forward/backward sweep algorithm for unbalanced power flow Power flow calculations are performed in the distribution system using the forward/backward sweep algorithm [22], which consists of an iterative series of voltage and current calculations used to determine the systems’ states. Considering split-phase secondary system components are represented as transfer matrices depicted in Fig. 4; equations for voltage (9) and current (10) are formulated in terms of the matrices ABCD.
Vjab = [A] Viab − [B] Iiab Ii
where Vi p is the per unit change in node i voltage at phase p between the present forward sweep value Vi p and the previp ous iteration value Vi,prev , considering nominal node phase voltage
5. Results and discussion
Transformer lifetime is limited by the winding of largest loading [20], which depends on the balancing of loads on the phases in the system. Given the case of balanced PBEV charging connections (50% of PBEV charge on split-phase A, 50% of PBEV charge on split-phase B), the annual expected transformer LOL is 2.24%; which is below the normal LOL limit of 5% (based on 20 years normal lifetime) [20]. Increasing the percentage of vehicles charging connected to split-phase A shows increased transformer aging, which exceeds the normal limit of 5% when more than 70% of PBEV use the same phase connection. Transformer lifetime degrades significantly faster in the case of 100% split-phase A charging (all PBEV charge on split-phase A), in which annual transformer LOL is more than double the normal limit; resulting in a transformer lifetime of less than 10 years.
ab
(11)
Vi,no min al .
4. Unbalanced loading on transformer loss of life
p
Vi,no min al
p
PPV,pu
p
||Vi | − |Vi,prev ||
= [C] Vjab + [D] Ijab
(9) (10)
where Vi ab and Ii ab denote the split-phase A and B voltages and currents at node i, and Vj ab and Ij ab denote the split-phase A and B voltages and currents at node j respectively. Forward voltage (9) and backward current (10) equations are repeated until the change in node voltages due to iterating (11) is within a specified tolerance, indicating the power systems state is converged. Upon convergence, the values for node voltages, line currents, and powers throughout the system are recorded. The resultant powers served by the transformers are then used for calculation of the transformer loss of life as given in Section 2.
Table 6 outlines the annual loss of life (LoL) experienced by the 50 kVA distribution transformer considering different PBEV and PV penetration rates as well as the vehicle charge starting time (VCST), with an equal distribution of PBEV charging on each phase of the secondary system. The annual loss of life experienced by the transformer is compared to the value of 5%, which corresponds to the normal lifetime of the transformer (i.e., 20 years) as per the IEEE standard C57.91-2011 [20]. Considering the base case of no PBEV (zero PBEV/house) and in the absence of rooftop solar PV (i.e., 0% PV) shown in Table 6, the transformer loss of life was found to be 2.4%. Considering normal transformer loss of life is taken as 5% annually, the loss of life of 2.4% is within acceptable limits. 5.1.1. Vehicles Charge Start Time (VCST) following home arrival time (HAT) without rooftop solar PV When PBEV begin to charge based on HAT through sampling the probability distribution function shown in Fig. 3, the increased loading of vehicles around peak residential loading times results in significant transformer loading, and consequently an annual LoL of 7.0%, exceeding the normal 5% LoL limit of the transformer. On inspection of the probability distribution function of HAT given in Fig. 3, it is seen that 6 p.m. is the peak HAT, and therefore most PBEV charging occurs in this time region. Furthermore, given that these afternoon hours also correspond with hours of high residential power consumption, the distribution transformer experiences significant loading and therefore reduced lifetime. 5.1.2. Vehicles Charge Start Time (VCST) following home arrival time (HAT) with rooftop solar PV Considering the case when PBEV begin to charge based on HAT and considering an active distribution system containing PV; the transformer annual LoL decreases from 7.0% without PV to 0.8% when all homes use 10 kW based on microFIT limits. As PV generation occurs during sunlight hours, this generation capability reduces the transformer loading during sunlight hours, which may include PBEV returning home prior to the evening hours of the day. Through the addition of PV in both cases of 50% and 100% penetration, the 5% loss of life limit corresponding to normal transformer lifetime of is not exceeded. As PBEV Level 1 charging is taken as 1.44 kW, and 50% PV penetration represents 5 kW of generation at each home; the generational capabilities of PV provide significant
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
207
Table 6 Yearly transformer loss of life in percent considering 50 kVA distribution transformer. Number of PBEV per house
Vehicle Charge Start Time (VCST)
0% PV
50% PV
100% PV
Zero One One One One
N/A HAT TOU-7 p.m. TOU-midnight Smart charging
2.4 7.0 7.6 4.3 4.3
0.5 1.7 2.4 1.1 1.0
0.3 0.8 1.4 0.7 0.6
reduction to the combined house and PBEV loading during sunlight hours. 5.1.3. Vehicles Charge Start Time (VCST) following time of use (TOU-7 p.m.) without rooftop solar PV When consumers take advantage of time of use pricing and charge simultaneously at 7 p.m., the resultant annual LoL seen on the transformer is 7.6%; which represents the VCST method resulting in the largest transformer loss of life and therefore the fastest transformer aging. Such increased loss of life is attributed to a combination of simultaneous PBEV charging at 7 p.m., which corresponds with peak loading times for residential homes. Furthermore, as 7.6% exceeds the 5% normal aging rate, the distribution transformer experienced reduced operational lifetime. 5.1.4. Vehicles Charge Start Time (VCST) following TOU-7 p.m. with rooftop solar PV In the case when all consumers begin charging starting at 7 p.m. to benefit from reduced electricity prices under TOU pricing and have 10 kW PV installed; the annual LoL of the transformer reduces to 1.4% versus 7.6% loss of life seen without 10 kW PV. Despite PV generation occurring sunlight hours, PV generation reduces residential loading during sunlight hours, reducing the hottest-spot temperature on the transformer, and consequently lowers annual LoL. Furthermore, it is seen that 50% PV penetration results in 2.4% annual LoL and therefore transformer LoL may be retained within normal 5% limits while supporting PBEV charging in the TOU-7 p.m. case when only some consumers in the system have PV installed. 5.1.5. Vehicles Charge Start Time (VCST) following TOU-midnight without rooftop solar PV When consumers delay charging their PBEV from 7 p.m. to midnight, not only they are able to take advantage of TOU pricing to reduce charging costs of the vehicles, but the increased system loading of PBEV is shifted to off-peak residential load times. As seen in Table 6; the violation of 5% transformer loss of life seen in simultaneous PBEV loading at 7 p.m. (TOU-7 p.m.) may be reduced through delaying simultaneous PBEV charging to midnight, in which case the resultant annual LoL on the transformer is 4.3% and falls within normal LoL limits. 5.1.6. Vehicles Charge Start Time (VCST) following TOU-midnight with rooftop solar PV While the case of PBEV charging beginning at midnight retains annual LoL within normal operation, the addition of rooftop PV allows for significant reduction in transformer annual LoL down to 0.7% with all consumers having 10 kW PV installed. 5.1.7. Vehicles Charge Start Time (VCST) following smart charging (SC) without rooftop solar PV In the case vehicles begin charging as two groups, the first of which starts at 11 p.m. with the second group starting at 3 a.m.; the resultant annual LoL on the transformer is 4.3%, which resides within the normal 5% value. Through splitting vehicles into two groups, the significant transformer loading seen when all vehicles charge simultaneously is reduced, and instead split into two smaller
Table 7 Peak transformer secondary winding neutral current for 0% PV and 1 PBEV/house. VCST
Time
Value
HAT TOU-7 p.m. TOU-midnight Smart charging Smart charging
7 p.m. 7 p.m. Midnight 11 p.m. 3 a.m.
19.3 A 30.8 A 30.2 A 22.0 A 25.7 A
portions. As the time delay between the charging blocks allows for some vehicles, which start charging at 11 p.m. to finish charging before the second group at 3 a.m. starts, the LoL seen by the transformer is comparable to that of the TOU-midnight method. 5.1.8. Vehicles Charge Start Time (VCST) following smart charging (SC) with rooftop solar PV The inclusion of rooftop solar PV combined with the 2-block smart charging method results in a minimum annual LoL of 0.6% when all homes have 10 kW PV installed. While TOU-midnight experiences 0.7% annual LoL at 100% PV penetration; vehicles which begin charging at 3 a.m. in the SC method with low state of charge may briefly overlap with the earlier sunrise hours of PV generation, resulting in reduced annual LoL on the transformer. 5.2. Center-tapped transformer neutral current Table 7 outlines the maximum expected neutral current seen at the secondary winding of the 50 kVA distribution transformer over the day given different VCST methods, including the time at which the peak neutral current was found to occur. As seen in Table 7, neutral current on the transformer is at a maximum when the largest number of PBEV are charging simultaneously; corresponding with 7 p.m. for TOU-7 p.m., midnight for TOU-midnight, and two peaks at 11 p.m. and 3 a.m. for the 2-block smart charging. As residential loads are assumed to be evenly divided on the split phases of the secondary system, and PV connections are 240 V and therefore are balanced among the split-phases; neutral current in the simulation only exists due to PBEV charging. 5.2.1. Vehicles Charge Start Time (VCST) following home arrival time (HAT) In the case vehicles begin to charge when returning home from work, as sampled from Fig. 3, the maximum neutral current seen on the transformer secondary is 19.3 A. Since most home owners would still try to take advantage of the reduced electricity rates according to TOU, VCST following HAT may not be popular and therefore the neutral current in case of VCST following TOU must be assessed. 5.2.2. Vehicles Charge Start Time (VCST) following time of use (TOU-7 p.m.) The case of TOU-7 p.m. VCST results in the largest transformer neutral current seen, with a value of 30.8 A occurring at 7 p.m. As PBEV charging 120 V are the only loads which may cause unbalance in the system, the maximum possible neutral current seen occurs when all vehicles charge simultaneously in the secondary circuit. As
208
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
TOU-7 p.m. experiences all vehicles charging at 7 p.m., it is expected that the neutral current takes on maximum value at 7 p.m. in this scenario. 5.2.3. Vehicles Charge Start Time (VCST) following time of use (TOU-midnight) When all vehicles charge simultaneously at midnight, the resultant neutral current seen in the transformer secondary is 30.2 A. While maximum neutral current is expected to occur when all vehicles charge simultaneously, the value of neutral current is slightly less than 30.8 A seen in TOU-7 p.m. As the number of PBEV on the system are identical, and PBEV are assumed constant power loads; the neutral current seen in TOU-midnight is less than TOU-7 p.m. as residential loading is smaller at midnight, which results in higher voltage magnitudes and consequently lower current values. 5.2.4. Vehicles Charge Start Time (VCST) following smart charging (SC) Given that smart charging partitions vehicle charging into two separate blocks (11 p.m. and 3 a.m.), the peak neutral currents of 22.0 A and 25.7 A at 11 p.m. and 3 a.m. respectively are shown in Table 7. Considering Level 1 charging results in potential charge times of over 4 h for vehicles which begin charging with a low state of charge, the peak neutral current experienced at 3 a.m. is larger than that at 11 p.m. Furthermore, smart charging experiences a maximum neutral current of 25.7 A versus 30.2 A seen in TOU-midnight, as separating two groups of vehicles charging in smart charging results in less PBEV charging simultaneously when compared with all PBEV in the system charging simultaneously at midnight in the TOU-midnight VCST. 6. Conclusion This paper investigates the effectiveness of PBEV charging schemes considering different Vehicle Charge Start Times and the synergistic effects of rooftop solar panel generation through the proposed methodology. The results of applying the proposed methodology to a 10-house secondary system have quantified the impact of PBEV charging on both loss of life and neutral current in distribution transformers. Results of the Monte Carlo simulation have shown that when PBEV are in the system, impact on transformer annual LoL is at a minimum when either TOU-midnight or SC VCST schemes are used (4.3% annual LoL). If PBEV charge when drivers return home from work or take advantage of time of use pricing immediately at 7 p.m. however; transformer loss of life experiences faster degradation than the expected lifetime (7.0% and 7.6% respectively). Transformer aging may be reduced through the addition of rooftop solar panel generation, which reduces transformer LoL for TOU-midnight and SC VCST methods by approximately 75% of the case without PV. Furthermore, the addition of 50% PV penetration is capable of reducing transformer LoL violations in HAT and TOU-7 p.m. charging methods to within acceptable design constraints. Considering TOU-midnight and smart charging schemes are the only vehicle charging methods that retain transformer annual LoL within normal limits; PBEV charging using the smart charging method results in a lower maximum neutral current than that of TOU-midnight (25.7 A versus 30.2 A respectively). As seen through applying the proposed methodology, PBEV charging results in an increase in both transformer aging and neutral current, an increase which is more pronounced when PBEV charge simultaneously (for such reasons as TOU pricing) versus more distributed methods (such as HAT). Furthermore, the addition of rooftop solar PV generation was found to decrease transformer loss of life; however, PV generation did not have an effect on reducing transformer neutral current. In this respect, it is recommended
that in order to mitigate the impact of PBEV charging on distribution companies, the Government of Ontario provide incentive for PBEV owners to charge on smart charging methods. Alternatively, continued incentive for rooftop solar PV installation may provide a different method for transformer aging reduction, however this does not relieve electric utilities experiencing neutral overload conditions. References [1] IESO, microFIT Program [Online]. Available: http://microfit.powerauthority. on.ca/about-microfit. [2] Ontario Ministry of Transportation, Cars are Evolving [Online]. Available: http://www.mto.gov.on.ca/english/dandv/vehicle/electric. [3] M.K. Gray, W.G. Morsi, Power quality assessment in distribution systems embedded with plug-in hybrid and battery electric vehicles, IEEE Trans. Power Syst. 30 (March (2)) (2015) 663–671. [4] H. Pezeshki, P.J. Wolfs, G. Ledwich, Impact of high PV penetration on distribution transformer insulation life, IEEE Trans. Power Deliv. 29 (June (3)) (2014) 1212–1220. [5] T.J. Geiles, S. Islam, Impact of PEV charging and rooftop PV penetration on distribution transformer life, in: Proc. IEEE PES-GM, Vancouver, BC, July 21–25, 2013, pp. 1–5. [6] M.S. ElNozahy, M.M.A. Salama, A comprehensive study of the impacts of PHEVs on residential distribution networks, IEEE Trans. Sustain. Energy 5 (January (1)) (2014) 332–342. [7] M.J. Rutherford, V. Yousefzadeh, The impact of electric vehicle battery charging on distribution transformers, in: Proc. IEEE APEC, Fort Worth, TX, March, 2011, pp. 396–400. [8] Q. Gong, S.M. Mohler, V. Marano, G. Rizzoni, PEV charging impact on residential distribution transformer life, in: IEEE EnergyTech, Cleveland, OH, May, 2011, pp. 1–6. [9] H. Turker, S. Bacha, D. Chatroux, A. Hably, Aging rate of low voltage transformer for a high penetration of plug-in hybrid electric vehicles (PHEVs), in: Proc. IEEE PES ISGT, Washington, DC, January, 2012, pp. 1–8, A. [10] K.J. Yunus, M. Reza, H.Z. Parra, K. Srivastava, Impacts of stochastic residential plug-in electric vehicle charging on distribution grid, in: Proc. IEEE PES ISGT, Washington, DC, January, 2012, pp. 1–8. [11] Q. Gong, S.M. Mohler, V. Marano, G. Rizzoni, Study of PEV charging on residential distribution transformer life, IEEE Trans. Smart Grid 3 (March (1)) (2012) 404–412. [12] H. Turker, S. Bacha, D. Chatroux, A. Hably, Low-voltage transformer loss-of-life assessments for a high penetration of plug-in hybrid electric vehicles (PHEVs), IEEE Trans. Power Deliv. 27 (July (3)) (2012) 1323–1331. [13] Q. Yan, M. Kezunovic, Impact analysis of electric vehicle charging on distribution system, in: Proc. IEEE NAPS, Champaign, IL, September, 2012, pp. 1–6. [14] R. Godina, N.G. Paterakis, O. Erdinc, E.M.G. Rodrigues, J.P.S. Catalao, Electric vehicles home charging impact on a distribution transformer in a Portuguese island, in: Proc. International Symposium on EDST, Vienna, Austria, September, 2015, pp. 74–79. [15] Y.O. Assolami, W.G. Morsi, Impact of second-generation plug-in battery electric vehicles on the aging of distribution transformers considering TOU prices, IEEE Trans. Sustain. Energy 6 (October (4)) (2015) 1606–1614. [16] G. Razeghi, L. Zhang, T. Brown, S. Samuelsen, Impacts of plug-in hybrid electric vehicles on a residential transformer using stochastic and empirical analysis, J. Power Sources 252 (December) (2013) 277–286. [17] K. Qian, C. Zhou, Y. Yuan, Impacts of high penetration level of fully electric vehicles charging loads on the thermal ageing of power transformers, Elect. Power Energy Syst. 65 (February) (2015) 102–112. [18] A.D. Hilshey, P.D.H. Hines, P. Rezaei, J.R. Dowds, Estimating the impact of electric vehicle smart charging on distribution transformer aging, IEEE Trans. Smart Grid 4 (June (2)) (2013) 905–913. [19] S.F. Abdelsamad, W.G. Morsi, T.S. Sidhu, Probabilistic impact of transportation electrification on the loss-of-life of distribution transformers in the presence of rooftop solar photovoltaic, IEEE Trans. Sustain. Energy 6 (October (4)) (2015) 1565–1573. [20] IEEE Guide for Loading Mineral-Oil-Immersed Transformers and Step-Voltage Regulators, IEEE Std. C57. 91-2011, New York, NY, USA, 2012 (March). [21] Distribution Test Feeders, IEEE Distribution System Analysis Subcommittee [Online]. Available: http://ewh.ieee.org/soc/pes/dsacom/testfeeders/. [22] W. Kersting, Distribution System Modeling and Analysis, 3rd ed., CRC Press, London, New York, 2012. [23] C.A. Burk, J.L. Bala, J.Z. Gibson, Electric secondary distribution system design, in: Proc. 39th IEEE North American Power Symposium (NAPS), Las Cruces, NM, October, 2007, pp. 582–588. [24] Reliability Test System Task Force, The IEEE reliability test system—1996, IEEE Trans. Power Syst. 14 (3) (1999) 1010–1020. [25] J. Bishop, Profiles on Residential Power Consumption, Final Report, Fire Protection Research Foundation, Quincy, MA, 2010 (March). [26] C. Jiang, R. Torquato, D. Salles, W. Xu, Method to assess the power-quality impact of plug-in electric vehicles, IEEE Trans. Power Deliv. 29 (2) (2014) 958–965.
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209 [27] National Household Travel Survey (NHTS) Data Center [Online]. Available: http:// nhts. ornl. gov/ download. shtml#2009. [28] SAE Electric Vehicle and Plug in Hybrid Electric Vehicle Conductive Charge Coupler, SAE J1772, 2010. [29] J.W. Smith, R. Dugan, W. Sunderman, Distribution modeling and analysis of high penetration PV, in: Proc. IEEE PES-GM, San Diego, CA, 2011, pp. 1–7. [30] Engineering Climate Datasets—Canadian Weather Energy and Engineering Datasets (CWEEDS) [Online]. Available: http://climate.weather.gc.ca/prods servs/engineering e.html.
209
[31] National Household Travel Survey Codebook [Online]. Available: http://nhts. ornl.gov/2009/pub/Codebook.pdf. [32] Ontario Energy Board, Electricity Prices—Time-of-use (TOU) Prices [Online], 2015 (May). Available: http://www.ontarioenergyboard.ca/OEB/Consumers/ Electricity/Electricity+Prices. [33] Ontario Power Authority, microFIT RULES Version 3.0, 2013 (August 28).
Contents lists available at ScienceDirect
Electric Power Systems Research journal homepage: www.elsevier.com/locate/epsr
On the impact of single-phase plug-in electric vehicles charging and rooftop solar photovoltaic on distribution transformer aging M.K. Gray, W.G. Morsi ∗ Faculty of Engineering and Applied Sciences, UOIT, Oshawa, ON, Canada
a r t i c l e
i n f o
Article history: Received 13 December 2016 Received in revised form 20 February 2017 Accepted 23 March 2017 Keywords: Monte Carlo methods Plug-in electric vehicles Power quality Rooftop solar photovoltaics Transformer aging
a b s t r a c t This study investigates the impact of single-phase plug-in electric vehicles charging on increasing the rate at which center-tapped distribution transformers experience aging. Distribution transformer aging is investigated considering varying rooftop solar photovoltaic generation penetration rates. Monte Carlo methods are used to probabilistically estimate the transformer’s loss of life considering the effect of timeof-use (TOU) pricing. The results of applying the proposed method have revealed that plug-in battery electric vehicle charging impact on both transformer aging and neutral current is largest in the case that vehicles charge based on time-of-use pricing methods. Further application has shown that while rooftop solar photovoltaic generation reduces transformer aging, no significant reduction in neutral current is observed. © 2017 Elsevier B.V. All rights reserved.
1. Introduction 1.1. Background In 2009, the Government of Ontario launched the micro feedin-tariff (microFIT) program [1], allowing homeowners to generate power from rooftop solar photovoltaic systems (PV) up to 10 kW, and get paid for the energy they produce over a 20-year term. Through adding such financial incentives for residential homeowners, the MicroFIT program aims to increase renewable energy generation in the province of Ontario. The Government of Ontario also launched the Ontario’s Electric Vehicles (OEV) incentive program and Green License Plates [2] in 2013, through which Ontario residents receive up to $8500 in rebates for the purchase or lease of new plug-in battery electric vehicles (PBEVs). 1.2. Problem statement While both microFIT and OEV incentives look to increase the penetration of Green technologies to reduce greenhouse gas (GHG) emissions in Ontario, the authors’ previous work [3] has shown that charging the battery of electric vehicles using Level 2 (240 V) charging may significantly increase distribution transformer loading for several hours, resulting in reduced transformer lifetimes. Studies [4,5] have shown increased rooftop solar PV penetration may
∗ Corresponding author at: University of Ontario Institute of Technology, Oshawa, ON L1H 7K4, Canada. Fax: +1 905 721 3370. E-mail address: [email protected] (W.G. Morsi). http://dx.doi.org/10.1016/j.epsr.2017.03.022 0378-7796/© 2017 Elsevier B.V. All rights reserved.
reduce loading on distribution transformers feeding the secondary system, and hence extending their lifetime; however these studies have not considered the split-phase nature of North American secondary systems, and therefore have not investigated the effects of rooftop solar generation considering Level 1 (120 V) plug-in electric vehicle charging. The unbalanced loading caused by electric vehicle charging at Level 1 (120 V) in residential homes has been investigated in Refs. [3,6]. The results of these studies have outlined increased neutral current and reduced distribution transformer’s lifetime as significant factors in considering split-phase electric vehicle charging impact. As the transformers lifetime is given as the split-phase winding which degrades the fastest, unbalanced loading due to PBEV charging reduces transformer lifetime faster than if the loading was balanced. In this respect, there is a need to quantify the effect of single-phase PBEV charging demand on the aging of distribution transformer’s life whilst considering rooftop solar PV generation.
1.3. Work to date While many works investigate the impact of electric vehicle charging on distribution transformer lifetime, a large number of studies [7–15] do not consider the effects of Level 1 electric vehicle charging; which charges using less power than Level 2 charging, for a longer duration of time. Studies considering Level 1 electric vehicle charging either do not consider the North American splitphase distribution transformer [16,17], or do not consider the effects of rooftop PV generation [3,6,18,19]. In this regard, no
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
203
Table 1 Transformer loss of life parameters.
Fig. 1. Center-tapped transformer connections to secondary system.
studies have investigated the mitigation potential of rooftop PV generation on reducing the impact caused by Level 1 electric vehicle charging. 1.4. Contribution
Parameter
Value
A H,R TO,R TO W m n R
30 ◦ C 27 ◦ C 53 ◦ C 6.86 h 0.08 h 0.8 0.8 4.87
Where A is the ambient temperature, H,R is the hottest-spot conductor rise above the top-oil temperature under rated load, TO,R is the top-oil rise above the ambient temperature under rated load, TO and W represent the top-oil and winding thermal time constants respectively, n and m are empirical top-oil rise and winding constants, and R is the ratio of load loss at the rated load to the no-load losses.
Table 2 Secondary distribution system extension nodes.
This paper presents a probabilistic approach using Monte Carlo to estimate the loss of life and neutral current impact on distribution transformers due to PBEV charging in an active distribution system. The power generation from rooftop solar photovoltaic is probabilistically estimated after modeling the secondary circuit components (e.g. transformer, service lines and service drops) feeding residential homes at which PBEVs charging is taking place. The approach presented in this study investigates the effect of time of use prices applied to electric vehicle charging as quantified in terms of distribution transformer aging and neutral current. Furthermore, the results of this study consider the impact reduction potential of rooftop solar PV penetration to reduce the impact of PBEV charging.
Node
Original load rating
Transformer rating
822 Phase A 846 Phase B 862 Phase B
50.69 kVA 17.00 kVA 20.87 kVA
50 kVA 25 kVA 25 kVA
Number of residential homes 10 6 6
2. Distribution transformer impact metrics 2.1. Distribution transformer neutral current Fig. 1 shows the equivalent circuit diagram of the center-tapped distribution transformer windings used to feed secondary circuits. When the currents drawn by split-phases A and B are not equal the presence of a neutral current IN may be observed. Kirchhoff’s Current Law (KCL) expresses this resultant transformer neutral current as the vector sum of the transformer split-phase currents in (1). In = Ia − Ib
(1)
Fig. 2. Ten house secondary system.
Transformer lifetime aging depends on the hottest-spot temperature, denoting the point of insulation on the transformer which deteriorates the fastest. Table 1 lists the thermal parameters needed for LoL calculations on a distribution transformer with 50 kVA nameplate rating [3].
where In , Ia , and Ib represent vectors of the neutral, split-phase A, and split-phase B currents measured at the transformer.
3. Probabilistic simulation methodology
2.2. Distribution transformer aging
3.1. Modified IEEE 34 Bus Test Distribution System
The IEEE Standard C57.91-2011 [20] details the standardized methodology used to calculate the percentage loss-of-life (LoL) of a given transformer based on the transformers hot-spot temperature.
Typically, distribution systems in North America consist of a primary system which extends secondary circuits used to feed residential customers. The primary system modeled in this work is taken as the IEEE 34 Bus Test Distribution System [21] using the exact lumped load model detailed in Ref. [22]. In order to accommodate secondary circuits, the load at each node listed in Table 2 (with apostrophes denoting intermediate node extension based on the lumped load model) was extended with a corresponding secondary circuit as follows. Single phase loads listed in Table 2 were removed and replaced with corresponding center-tapped transformers as rated in Table 2. Each transformer was further extended with a secondary circuit used to feed homes, with circuit model based on the archetype design in Ref. [23]. An example layout of the 10-house secondary distribution system (SDS) fed from a 50 kVA distribution transformer is depicted in Fig. 2 with residential house nodes labeled nH1 to nH10 . As detailed in Ref. [23], secondary lines (SL) and service drops (SD) are assumed 4/0 AA and 1/0 AA triplex cables respectively.
LoL(%) =
FEQA × t × 100 Normal Insulation Life
(2)
where the normal insulation life of a transformer is typically 180,000 h [20], t is the total number of hours in a day, and FEQA is the average aging factor. FEQA =
N
FAA,n tn
N /
n=1
tn
(3)
n=1
Given time step size tn , the aging acceleration factor FAA,n determines the increase in transformer lifetime degradation based on the winding hottest-spot temperature H . FAA = e
15000 383
−
15000 H +273
(4)
204
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
Table 3 Vehicle and charger parameters.
Table 4 Temperature factor table.
Parameter
Value
T (◦ C)
FT (T)
εv Cv SOCmin c
0.24 kWh/mile 24 kWh 5% 0.90
0 25 75 100
1.2 1.0 0.8 0.6
3.2. Residential load profiles The IEEE Reliability Test System (RTS) [24] details a representative load profile given as a percent of the peak load at an hourly resolution over a year. Through sampling the day of the year with uniform random probability, the daily load profile for each residential home in the secondary is obtained from the IEEE RTS, assuming a peak load of 6.64 kVA with 0.9 lagging power factor [25]. Residential loads are assumed constant power loads, with hourly power expressed as: Sn (h) = SnP · Pn (h)
(5) SP
where Sn (h) is the apparent power of house n at hour h, n is the peak apparent power of house n, and Pn (h) is the percent of peak load experienced by house n at hour h for the day sampled from the IEEE RTS.
depends on the rating of the PV array PPV,Rated in kW, the solar irradiance Irad in kW/m2 and the corresponding temperature factor FT (T) for temperature T linearly interpolated from Table 4 (8). PPV,Array = PPV,Rated × Irad × FT (T )
(8)
The power output from rooftop solar photovoltaics (PV’s) to the alternating current (AC) grid PPV,Out is calculated by considering the inverters efficiency Inverter applied to the direct current (DC) power generated from the array. Inverter efficiency is calculated based on the per unit output of the PV array in Table 5 [29]. Hourly irradiance and temperature seen by each rooftop solar PV are taken from the Canadian Weather Energy and Engineering Datasets (CWEEDS) 2005 data in Toronto, Ontario [30] based on the uniform randomly sampled day of the year. The number of PVs in the system is as described in Section 3.5. 3.5. Monte Carlo algorithm
3.3. Plug-in battery electric vehicles charging demand Plug-in battery electric vehicle charging demand is estimated in this study using the model developed in Ref. [26], which calculates total electric vehicle charging energy considering the state-ofcharge (SOC) of the electric vehicle’s battery and the efficiency of the electric vehicle charger. The SOC of the electric vehicle depends on the distance the electric vehicle has driven as well as the electric vehicle’s mileage in terms of energy consumption. Vehicle driving distance is sampled from the National Household Travel Survey (NHTS) 2009 [27] which reports annual mileage (ANNMILES) of surveyed vehicles. The daily distances traveled were obtained by normalizing the annual mileage through division by 365 representing the days in a year. The total energy required by each vehicle v to fully charge is then calculated using (6) [3].
Ev =
max dv × εv , Ev,min
C
(6)
With Ev the total energy required to fully charge vehicle v, dv the sampled driving distance (miles), εv the energy consumption (kWh/mile), C the charger efficiency (per unit), and Ev, min the largest amount of energy required by the vehicle when the battery is at minimum SOC calculable in (7) [3]. Ev,min = Cv × SOCmin
A Monte Carlo simulation is used in this work to address the uncertainties associated with PBEV charging (e.g., daily driving distances, charging time, etc.) and rooftop solar PV generation (e.g., daily temperature and solar irradiance). Through performing a number of independent random experiments (Monte Carlo trials) the impact metrics are taken as the expected value of the statistical results. The simulation in this work considers two PBEV penetrations: no PBEV in the system and one PBEV at each house. Driving distance for each PBEV in the system is sampled using NHTS outlined in Section 3.3, which begin charging based on the Vehicle Charge Start Time (VCST). Simulations are performed for four separate VCST schemes: home arrival time (HAT), time of use 7 p.m. (TOU-7 P.M.), time of use midnight (TOU-midnight), and smart charging (SC) as described in the following subsections.
(7)
where Cv is the battery capacity of the vehicle and SOCmin is the minimum state of charge of the vehicle in per unit. All vehicles in this study are assumed Nissan Leaf, with vehicle and charger parameters outlined in Table 3. Electric vehicle charging on the distribution system further assumes a constant power charging model at unity power factor, using Level 1120 V charging rated 1.44 kW based on SAE J1772 electric vehicle charging standards [28]. The number of vehicles in the system is varied as outlined in Section 3.5. 3.4. Rooftop solar PV power generation model Residential rooftop solar PV installations consist of a solar array, which generates DC power, the power of which is then passed through an inverter to connect to the AC distribution system. The DC power generated from residential rooftop PV arrays PPV,Array
3.5.1. Home arrival time (HAT) Home arrival time charging assumes each PBEV owner will begin to charge their vehicle based on the time at which the PBEV owner returns home from work. The probability distribution for HAT charging is derived from the National Household Travel Survey (NHTS) [27] as follows. Following the NHTS Codebook [31], the subset of vehicles flagged as returning home from work are extracted from the original dataset, and a probability distribution is created from the vehicle trip “ENDTIME” to determine the hour at which a vehicle begins charging. The resultant probability density function (PDF) for vehicles home arrival times is depicted in Fig. 3, with the most common hour of arrival at 6 p.m. (i.e., hour 18 of the day). In the case that PBEV charging begins based on HAT, the home arrival time PDF is sampled for each PBEV in the system to determine when the PBEV begins charging. 3.5.2. Time of use (TOU) 7 p.m. Time of use charging reflects the notion that consumers will strive to minimize the cost of vehicle charging, by taking advantage of the reduced electricity rates according to time of use (TOU) in Ontario, Canada [32]. In this respect, the start of off-peak pricing begins at 7 p.m., in which all vehicles are assumed to begin charging simultaneously.
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
3.5.3. Time of use (TOU) midnight While TOU off-peak pricing begins at 7 p.m., this time coincides with peak loading of residential homes. In order for electric vehicle owners to maintain low electricity prices to charge vehicles and reduce the risk of overloading the transformer; TOU-midnight assumes a hypothetical scenario in which off-peak prices for PBEV charging begin at midnight. In this respect, all vehicles begin charging at midnight, a time at which residential loading is off-peak. 3.5.4. Smart charging The smart charging scheme aims to minimize the expectedly higher impact of PBEV charging on the distribution system simultaneously by splitting vehicle charging equally into two groups [15]. In this respect, the first group of PBEV on the system begin charging at 11 p.m., whereas the second group of PBEV begin to charge starting at 3 a.m. Through splitting vehicle charging into two groups, the SC approach aims to allow customers to charge vehicles during the off-peak TOU pricing, with reduced impact on the distribution system.
Rooftop solar PV arrays are assigned equally to all households from 0 kW/house to 10 kW/house based on MicroFIT limits [33], with a step size of 5 kW. The percentage PV penetration varies from 0% (i.e., 0 kW/House) to 100% (10 kW/House). Hourly temperature and irradiance data are sampled for each Monte Carlo trial based on CWEEDS data outlined in Section 3.4. House load profiles are sampled once for each secondary system of a Monte Carlo trial using IEEE RTS as outlined in Section 3.2. The Monte Carlo simulation is further detailed in Algorithm 1. The number of Monte Carlo trials required by the simulation to converge has been found to be 6000 following visual inspection based on mean transformer LOL, through which additional Monte Carlo trials resulted in negligible change to mean annual transformer LOL.
Algorithm 1.
Monte Carlo simulation procedure
1: Start Input parameters: Home arrival time and daily driving distance CDFs Temperature and irradiance CDF’s Nissan Leaf Parameters (εv,Cv,SOCmin,ηC) Number of Monte Carlo Trials (NMC) 2: For trial = 1:1:NMC 3:
For PBEV/House = 0:1
4:
For each vehicle v
5:
Assign vehicle to house
6:
Determine VCSTv (HAT, TOU-7pm, or TOU-midnight)
7:
Sample dv from daily driving distance
8: 9:
Calculate Ev using (6) End for
10: Assign kW rating to PV at each house 11: Sample daily temperature and irradiance at hourly resolution using CWEEDS 2005 data in Toronto, Ontario 12:
For t=1:24
13:
Perform backward/forward sweep
14:
Record neutral current and center-tapped distribution transformer annual loss of life
15: End for 16: End for
17: End for 18: Evaluate Monte Carlo data 19: End
205
206
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
Fig. 4. Split-phase component model.
Forward/backward sweep is repeated for each hour of the day as specified previously in Algorithm 1. p
p
Vi (pu) = Fig. 3. Probability density function for sampling the hour at which a PBEV arrives home from work (i.e. begins to charge). Table 5 Inverter efficiency table. FInverter (PPV,pu )
0.1 0.2 0.4 1.0
0.86 0.90 0.93 0.97
5.1. Transformer loss-of-life
4.1. Forward/backward sweep algorithm for unbalanced power flow Power flow calculations are performed in the distribution system using the forward/backward sweep algorithm [22], which consists of an iterative series of voltage and current calculations used to determine the systems’ states. Considering split-phase secondary system components are represented as transfer matrices depicted in Fig. 4; equations for voltage (9) and current (10) are formulated in terms of the matrices ABCD.
Vjab = [A] Viab − [B] Iiab Ii
where Vi p is the per unit change in node i voltage at phase p between the present forward sweep value Vi p and the previp ous iteration value Vi,prev , considering nominal node phase voltage
5. Results and discussion
Transformer lifetime is limited by the winding of largest loading [20], which depends on the balancing of loads on the phases in the system. Given the case of balanced PBEV charging connections (50% of PBEV charge on split-phase A, 50% of PBEV charge on split-phase B), the annual expected transformer LOL is 2.24%; which is below the normal LOL limit of 5% (based on 20 years normal lifetime) [20]. Increasing the percentage of vehicles charging connected to split-phase A shows increased transformer aging, which exceeds the normal limit of 5% when more than 70% of PBEV use the same phase connection. Transformer lifetime degrades significantly faster in the case of 100% split-phase A charging (all PBEV charge on split-phase A), in which annual transformer LOL is more than double the normal limit; resulting in a transformer lifetime of less than 10 years.
ab
(11)
Vi,no min al .
4. Unbalanced loading on transformer loss of life
p
Vi,no min al
p
PPV,pu
p
||Vi | − |Vi,prev ||
= [C] Vjab + [D] Ijab
(9) (10)
where Vi ab and Ii ab denote the split-phase A and B voltages and currents at node i, and Vj ab and Ij ab denote the split-phase A and B voltages and currents at node j respectively. Forward voltage (9) and backward current (10) equations are repeated until the change in node voltages due to iterating (11) is within a specified tolerance, indicating the power systems state is converged. Upon convergence, the values for node voltages, line currents, and powers throughout the system are recorded. The resultant powers served by the transformers are then used for calculation of the transformer loss of life as given in Section 2.
Table 6 outlines the annual loss of life (LoL) experienced by the 50 kVA distribution transformer considering different PBEV and PV penetration rates as well as the vehicle charge starting time (VCST), with an equal distribution of PBEV charging on each phase of the secondary system. The annual loss of life experienced by the transformer is compared to the value of 5%, which corresponds to the normal lifetime of the transformer (i.e., 20 years) as per the IEEE standard C57.91-2011 [20]. Considering the base case of no PBEV (zero PBEV/house) and in the absence of rooftop solar PV (i.e., 0% PV) shown in Table 6, the transformer loss of life was found to be 2.4%. Considering normal transformer loss of life is taken as 5% annually, the loss of life of 2.4% is within acceptable limits. 5.1.1. Vehicles Charge Start Time (VCST) following home arrival time (HAT) without rooftop solar PV When PBEV begin to charge based on HAT through sampling the probability distribution function shown in Fig. 3, the increased loading of vehicles around peak residential loading times results in significant transformer loading, and consequently an annual LoL of 7.0%, exceeding the normal 5% LoL limit of the transformer. On inspection of the probability distribution function of HAT given in Fig. 3, it is seen that 6 p.m. is the peak HAT, and therefore most PBEV charging occurs in this time region. Furthermore, given that these afternoon hours also correspond with hours of high residential power consumption, the distribution transformer experiences significant loading and therefore reduced lifetime. 5.1.2. Vehicles Charge Start Time (VCST) following home arrival time (HAT) with rooftop solar PV Considering the case when PBEV begin to charge based on HAT and considering an active distribution system containing PV; the transformer annual LoL decreases from 7.0% without PV to 0.8% when all homes use 10 kW based on microFIT limits. As PV generation occurs during sunlight hours, this generation capability reduces the transformer loading during sunlight hours, which may include PBEV returning home prior to the evening hours of the day. Through the addition of PV in both cases of 50% and 100% penetration, the 5% loss of life limit corresponding to normal transformer lifetime of is not exceeded. As PBEV Level 1 charging is taken as 1.44 kW, and 50% PV penetration represents 5 kW of generation at each home; the generational capabilities of PV provide significant
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
207
Table 6 Yearly transformer loss of life in percent considering 50 kVA distribution transformer. Number of PBEV per house
Vehicle Charge Start Time (VCST)
0% PV
50% PV
100% PV
Zero One One One One
N/A HAT TOU-7 p.m. TOU-midnight Smart charging
2.4 7.0 7.6 4.3 4.3
0.5 1.7 2.4 1.1 1.0
0.3 0.8 1.4 0.7 0.6
reduction to the combined house and PBEV loading during sunlight hours. 5.1.3. Vehicles Charge Start Time (VCST) following time of use (TOU-7 p.m.) without rooftop solar PV When consumers take advantage of time of use pricing and charge simultaneously at 7 p.m., the resultant annual LoL seen on the transformer is 7.6%; which represents the VCST method resulting in the largest transformer loss of life and therefore the fastest transformer aging. Such increased loss of life is attributed to a combination of simultaneous PBEV charging at 7 p.m., which corresponds with peak loading times for residential homes. Furthermore, as 7.6% exceeds the 5% normal aging rate, the distribution transformer experienced reduced operational lifetime. 5.1.4. Vehicles Charge Start Time (VCST) following TOU-7 p.m. with rooftop solar PV In the case when all consumers begin charging starting at 7 p.m. to benefit from reduced electricity prices under TOU pricing and have 10 kW PV installed; the annual LoL of the transformer reduces to 1.4% versus 7.6% loss of life seen without 10 kW PV. Despite PV generation occurring sunlight hours, PV generation reduces residential loading during sunlight hours, reducing the hottest-spot temperature on the transformer, and consequently lowers annual LoL. Furthermore, it is seen that 50% PV penetration results in 2.4% annual LoL and therefore transformer LoL may be retained within normal 5% limits while supporting PBEV charging in the TOU-7 p.m. case when only some consumers in the system have PV installed. 5.1.5. Vehicles Charge Start Time (VCST) following TOU-midnight without rooftop solar PV When consumers delay charging their PBEV from 7 p.m. to midnight, not only they are able to take advantage of TOU pricing to reduce charging costs of the vehicles, but the increased system loading of PBEV is shifted to off-peak residential load times. As seen in Table 6; the violation of 5% transformer loss of life seen in simultaneous PBEV loading at 7 p.m. (TOU-7 p.m.) may be reduced through delaying simultaneous PBEV charging to midnight, in which case the resultant annual LoL on the transformer is 4.3% and falls within normal LoL limits. 5.1.6. Vehicles Charge Start Time (VCST) following TOU-midnight with rooftop solar PV While the case of PBEV charging beginning at midnight retains annual LoL within normal operation, the addition of rooftop PV allows for significant reduction in transformer annual LoL down to 0.7% with all consumers having 10 kW PV installed. 5.1.7. Vehicles Charge Start Time (VCST) following smart charging (SC) without rooftop solar PV In the case vehicles begin charging as two groups, the first of which starts at 11 p.m. with the second group starting at 3 a.m.; the resultant annual LoL on the transformer is 4.3%, which resides within the normal 5% value. Through splitting vehicles into two groups, the significant transformer loading seen when all vehicles charge simultaneously is reduced, and instead split into two smaller
Table 7 Peak transformer secondary winding neutral current for 0% PV and 1 PBEV/house. VCST
Time
Value
HAT TOU-7 p.m. TOU-midnight Smart charging Smart charging
7 p.m. 7 p.m. Midnight 11 p.m. 3 a.m.
19.3 A 30.8 A 30.2 A 22.0 A 25.7 A
portions. As the time delay between the charging blocks allows for some vehicles, which start charging at 11 p.m. to finish charging before the second group at 3 a.m. starts, the LoL seen by the transformer is comparable to that of the TOU-midnight method. 5.1.8. Vehicles Charge Start Time (VCST) following smart charging (SC) with rooftop solar PV The inclusion of rooftop solar PV combined with the 2-block smart charging method results in a minimum annual LoL of 0.6% when all homes have 10 kW PV installed. While TOU-midnight experiences 0.7% annual LoL at 100% PV penetration; vehicles which begin charging at 3 a.m. in the SC method with low state of charge may briefly overlap with the earlier sunrise hours of PV generation, resulting in reduced annual LoL on the transformer. 5.2. Center-tapped transformer neutral current Table 7 outlines the maximum expected neutral current seen at the secondary winding of the 50 kVA distribution transformer over the day given different VCST methods, including the time at which the peak neutral current was found to occur. As seen in Table 7, neutral current on the transformer is at a maximum when the largest number of PBEV are charging simultaneously; corresponding with 7 p.m. for TOU-7 p.m., midnight for TOU-midnight, and two peaks at 11 p.m. and 3 a.m. for the 2-block smart charging. As residential loads are assumed to be evenly divided on the split phases of the secondary system, and PV connections are 240 V and therefore are balanced among the split-phases; neutral current in the simulation only exists due to PBEV charging. 5.2.1. Vehicles Charge Start Time (VCST) following home arrival time (HAT) In the case vehicles begin to charge when returning home from work, as sampled from Fig. 3, the maximum neutral current seen on the transformer secondary is 19.3 A. Since most home owners would still try to take advantage of the reduced electricity rates according to TOU, VCST following HAT may not be popular and therefore the neutral current in case of VCST following TOU must be assessed. 5.2.2. Vehicles Charge Start Time (VCST) following time of use (TOU-7 p.m.) The case of TOU-7 p.m. VCST results in the largest transformer neutral current seen, with a value of 30.8 A occurring at 7 p.m. As PBEV charging 120 V are the only loads which may cause unbalance in the system, the maximum possible neutral current seen occurs when all vehicles charge simultaneously in the secondary circuit. As
208
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209
TOU-7 p.m. experiences all vehicles charging at 7 p.m., it is expected that the neutral current takes on maximum value at 7 p.m. in this scenario. 5.2.3. Vehicles Charge Start Time (VCST) following time of use (TOU-midnight) When all vehicles charge simultaneously at midnight, the resultant neutral current seen in the transformer secondary is 30.2 A. While maximum neutral current is expected to occur when all vehicles charge simultaneously, the value of neutral current is slightly less than 30.8 A seen in TOU-7 p.m. As the number of PBEV on the system are identical, and PBEV are assumed constant power loads; the neutral current seen in TOU-midnight is less than TOU-7 p.m. as residential loading is smaller at midnight, which results in higher voltage magnitudes and consequently lower current values. 5.2.4. Vehicles Charge Start Time (VCST) following smart charging (SC) Given that smart charging partitions vehicle charging into two separate blocks (11 p.m. and 3 a.m.), the peak neutral currents of 22.0 A and 25.7 A at 11 p.m. and 3 a.m. respectively are shown in Table 7. Considering Level 1 charging results in potential charge times of over 4 h for vehicles which begin charging with a low state of charge, the peak neutral current experienced at 3 a.m. is larger than that at 11 p.m. Furthermore, smart charging experiences a maximum neutral current of 25.7 A versus 30.2 A seen in TOU-midnight, as separating two groups of vehicles charging in smart charging results in less PBEV charging simultaneously when compared with all PBEV in the system charging simultaneously at midnight in the TOU-midnight VCST. 6. Conclusion This paper investigates the effectiveness of PBEV charging schemes considering different Vehicle Charge Start Times and the synergistic effects of rooftop solar panel generation through the proposed methodology. The results of applying the proposed methodology to a 10-house secondary system have quantified the impact of PBEV charging on both loss of life and neutral current in distribution transformers. Results of the Monte Carlo simulation have shown that when PBEV are in the system, impact on transformer annual LoL is at a minimum when either TOU-midnight or SC VCST schemes are used (4.3% annual LoL). If PBEV charge when drivers return home from work or take advantage of time of use pricing immediately at 7 p.m. however; transformer loss of life experiences faster degradation than the expected lifetime (7.0% and 7.6% respectively). Transformer aging may be reduced through the addition of rooftop solar panel generation, which reduces transformer LoL for TOU-midnight and SC VCST methods by approximately 75% of the case without PV. Furthermore, the addition of 50% PV penetration is capable of reducing transformer LoL violations in HAT and TOU-7 p.m. charging methods to within acceptable design constraints. Considering TOU-midnight and smart charging schemes are the only vehicle charging methods that retain transformer annual LoL within normal limits; PBEV charging using the smart charging method results in a lower maximum neutral current than that of TOU-midnight (25.7 A versus 30.2 A respectively). As seen through applying the proposed methodology, PBEV charging results in an increase in both transformer aging and neutral current, an increase which is more pronounced when PBEV charge simultaneously (for such reasons as TOU pricing) versus more distributed methods (such as HAT). Furthermore, the addition of rooftop solar PV generation was found to decrease transformer loss of life; however, PV generation did not have an effect on reducing transformer neutral current. In this respect, it is recommended
that in order to mitigate the impact of PBEV charging on distribution companies, the Government of Ontario provide incentive for PBEV owners to charge on smart charging methods. Alternatively, continued incentive for rooftop solar PV installation may provide a different method for transformer aging reduction, however this does not relieve electric utilities experiencing neutral overload conditions. References [1] IESO, microFIT Program [Online]. Available: http://microfit.powerauthority. on.ca/about-microfit. [2] Ontario Ministry of Transportation, Cars are Evolving [Online]. Available: http://www.mto.gov.on.ca/english/dandv/vehicle/electric. [3] M.K. Gray, W.G. Morsi, Power quality assessment in distribution systems embedded with plug-in hybrid and battery electric vehicles, IEEE Trans. Power Syst. 30 (March (2)) (2015) 663–671. [4] H. Pezeshki, P.J. Wolfs, G. Ledwich, Impact of high PV penetration on distribution transformer insulation life, IEEE Trans. Power Deliv. 29 (June (3)) (2014) 1212–1220. [5] T.J. Geiles, S. Islam, Impact of PEV charging and rooftop PV penetration on distribution transformer life, in: Proc. IEEE PES-GM, Vancouver, BC, July 21–25, 2013, pp. 1–5. [6] M.S. ElNozahy, M.M.A. Salama, A comprehensive study of the impacts of PHEVs on residential distribution networks, IEEE Trans. Sustain. Energy 5 (January (1)) (2014) 332–342. [7] M.J. Rutherford, V. Yousefzadeh, The impact of electric vehicle battery charging on distribution transformers, in: Proc. IEEE APEC, Fort Worth, TX, March, 2011, pp. 396–400. [8] Q. Gong, S.M. Mohler, V. Marano, G. Rizzoni, PEV charging impact on residential distribution transformer life, in: IEEE EnergyTech, Cleveland, OH, May, 2011, pp. 1–6. [9] H. Turker, S. Bacha, D. Chatroux, A. Hably, Aging rate of low voltage transformer for a high penetration of plug-in hybrid electric vehicles (PHEVs), in: Proc. IEEE PES ISGT, Washington, DC, January, 2012, pp. 1–8, A. [10] K.J. Yunus, M. Reza, H.Z. Parra, K. Srivastava, Impacts of stochastic residential plug-in electric vehicle charging on distribution grid, in: Proc. IEEE PES ISGT, Washington, DC, January, 2012, pp. 1–8. [11] Q. Gong, S.M. Mohler, V. Marano, G. Rizzoni, Study of PEV charging on residential distribution transformer life, IEEE Trans. Smart Grid 3 (March (1)) (2012) 404–412. [12] H. Turker, S. Bacha, D. Chatroux, A. Hably, Low-voltage transformer loss-of-life assessments for a high penetration of plug-in hybrid electric vehicles (PHEVs), IEEE Trans. Power Deliv. 27 (July (3)) (2012) 1323–1331. [13] Q. Yan, M. Kezunovic, Impact analysis of electric vehicle charging on distribution system, in: Proc. IEEE NAPS, Champaign, IL, September, 2012, pp. 1–6. [14] R. Godina, N.G. Paterakis, O. Erdinc, E.M.G. Rodrigues, J.P.S. Catalao, Electric vehicles home charging impact on a distribution transformer in a Portuguese island, in: Proc. International Symposium on EDST, Vienna, Austria, September, 2015, pp. 74–79. [15] Y.O. Assolami, W.G. Morsi, Impact of second-generation plug-in battery electric vehicles on the aging of distribution transformers considering TOU prices, IEEE Trans. Sustain. Energy 6 (October (4)) (2015) 1606–1614. [16] G. Razeghi, L. Zhang, T. Brown, S. Samuelsen, Impacts of plug-in hybrid electric vehicles on a residential transformer using stochastic and empirical analysis, J. Power Sources 252 (December) (2013) 277–286. [17] K. Qian, C. Zhou, Y. Yuan, Impacts of high penetration level of fully electric vehicles charging loads on the thermal ageing of power transformers, Elect. Power Energy Syst. 65 (February) (2015) 102–112. [18] A.D. Hilshey, P.D.H. Hines, P. Rezaei, J.R. Dowds, Estimating the impact of electric vehicle smart charging on distribution transformer aging, IEEE Trans. Smart Grid 4 (June (2)) (2013) 905–913. [19] S.F. Abdelsamad, W.G. Morsi, T.S. Sidhu, Probabilistic impact of transportation electrification on the loss-of-life of distribution transformers in the presence of rooftop solar photovoltaic, IEEE Trans. Sustain. Energy 6 (October (4)) (2015) 1565–1573. [20] IEEE Guide for Loading Mineral-Oil-Immersed Transformers and Step-Voltage Regulators, IEEE Std. C57. 91-2011, New York, NY, USA, 2012 (March). [21] Distribution Test Feeders, IEEE Distribution System Analysis Subcommittee [Online]. Available: http://ewh.ieee.org/soc/pes/dsacom/testfeeders/. [22] W. Kersting, Distribution System Modeling and Analysis, 3rd ed., CRC Press, London, New York, 2012. [23] C.A. Burk, J.L. Bala, J.Z. Gibson, Electric secondary distribution system design, in: Proc. 39th IEEE North American Power Symposium (NAPS), Las Cruces, NM, October, 2007, pp. 582–588. [24] Reliability Test System Task Force, The IEEE reliability test system—1996, IEEE Trans. Power Syst. 14 (3) (1999) 1010–1020. [25] J. Bishop, Profiles on Residential Power Consumption, Final Report, Fire Protection Research Foundation, Quincy, MA, 2010 (March). [26] C. Jiang, R. Torquato, D. Salles, W. Xu, Method to assess the power-quality impact of plug-in electric vehicles, IEEE Trans. Power Deliv. 29 (2) (2014) 958–965.
M.K. Gray, W.G. Morsi / Electric Power Systems Research 148 (2017) 202–209 [27] National Household Travel Survey (NHTS) Data Center [Online]. Available: http:// nhts. ornl. gov/ download. shtml#2009. [28] SAE Electric Vehicle and Plug in Hybrid Electric Vehicle Conductive Charge Coupler, SAE J1772, 2010. [29] J.W. Smith, R. Dugan, W. Sunderman, Distribution modeling and analysis of high penetration PV, in: Proc. IEEE PES-GM, San Diego, CA, 2011, pp. 1–7. [30] Engineering Climate Datasets—Canadian Weather Energy and Engineering Datasets (CWEEDS) [Online]. Available: http://climate.weather.gc.ca/prods servs/engineering e.html.
209
[31] National Household Travel Survey Codebook [Online]. Available: http://nhts. ornl.gov/2009/pub/Codebook.pdf. [32] Ontario Energy Board, Electricity Prices—Time-of-use (TOU) Prices [Online], 2015 (May). Available: http://www.ontarioenergyboard.ca/OEB/Consumers/ Electricity/Electricity+Prices. [33] Ontario Power Authority, microFIT RULES Version 3.0, 2013 (August 28).