The effects of climate and climate change on electric vehicle charging demand in Toronto, Canada

Battery electric vehicles (BEVs) influence total and peak electricity demand, but few studies account for climate when studying these effects. This study quantifies BEV charging demand in the Greater Toronto and Hamilton Area using a detailed trip level approach, accounting for the effect of present and future temperatures on BEV energy consumption. The impact of temperature on charging demand was largest in winter. In 2019, charging demand increases by 52% on an average January day, and up to 82% on extreme days (relative to mild weather conditions). At 30% penetration, BEVs increase peak demand on January’s coldest day by 600–3600 MW (3%–5%), of which 300–700 MW is driven by temperature, depending on the charging scenario. Climate change introduces small changes, increasing summer and decreasing winter charging demand. These results highlight the importance of adjusting for regional climate variation and temperature extremes when analyzing the impact of BEVs on the grid.


Introduction
The Canadian transportation sector emitted 186 Mt of CO 2 eq in 2019, roughly one quarter of national greenhouse gas (GHG) emissions (Environment and Climate Change Canada 2021a).Given the increasing concern of GHG emissions contributing to anthropogenic climate change, governments throughout the world are establishing action plans to mitigate or reduce road transport emissions.For example, the Government of Canada has set a target for 100% of light duty vehicle sales to be zero emission vehicles by 2035 (Transport Canada 2021).Electric vehicles (EVs), including battery electric vehicles (BEVs) and plug-in hybrid vehicles (PHEVs), are a key component of GHG emissions reduction in the transportation sector under the International Energy Agency's Net Zero by 2050 scenario (International Energy Agency 2021a).
With the rising rate of EV adoption, concerns arise regarding the ability of the existing electrical grid to meet future EV charging demand.It is imperative that policymakers can forecast a probable charge profile and manage the additional electricity demand to maintain the stability of the grid.Consequently, the potential for EV charging demand to have significant impacts on the existing electricity system has been recognized by system operators (e.g.Independent Electricity System Operator 2020a, Alberta Utilities Commission 2021) as well as academia (e.g. Green II et al 2011, Foley et al 2013, Muratori 2018, Gryparis et al 2020).The location and timing of EV charging is also crucial.The Independent Electricity System Operator (IESO), the organization responsible for operating Ontario's electricity system, indicated in its 2020 Annual Planning Outlook Report that the charging profile of EVs is as influential as the total EV charging energy when considering the effects on the demand forecast (Independent Electricity System Operator 2020a).
However, the energy consumption rate of EVs is dependent on ambient temperature, which means that climate and in turn climate change can influence the energy consumption of EVs.Increases to energy consumption and respective decreases to driving range have been observed in both hot and cold temperatures (Lohse-Busch et al 2013, Yuksel andMichalek 2015, American Automobile Association 2019).The regional heterogeneity of EV electricity demand due to regional climate has also been illustrated in works studying EV GHG emissions, e.g.Yuksel and Michalek (2015).As such, there is potential for climate and climate change to materially affect the impacts of EV charging on the electric grid.
The influence of ambient temperature on EV impacts has been explored in the literature, primarily in the context of GHG emissions.Miller et al (2020) explored the influence of temperature among other factors on BEV lifecycle GHG emissions.They also conducted a literature review of studies in that space, finding only 10% of them consider the influence of temperature.Research in this space has shown there is regional heterogeneity for EV GHG emissions, with disparities caused by an area's grid mix, driving patterns, driving conditions, as well as temperature (Yuksel and Michalek 2015, Hoehne and Chester 2016, Wu et al 2019, Miller et al 2020).Battery performance and lifespan are also of interest since EV lifecycle GHG emissions are influenced by battery production (e.g.Notter et al 2010, Hawkins et al 2013, Elgowainy et al 2018).Studies have considered the effects of ambient temperature on charging and discharging batteries, as well as analyzing battery degradation while factoring in temperature changes (Archsmith et al 2015, Yang et al 2018).
Temperature has been considered to a limited extent in analyses of BEV charging impacts on the grid.A literature review of EV load modelling from Amara-Ouali et al (2021) noted consideration of temperature was rare in EV load modelling, even though it is commonplace in electric load modelling.Consideration of temperature in BEV charging impacts has been done through stochastic models with a range of BEV energy consumption rates (Soares et al 2011, Wu et al 2011) or empirical data capturing BEV energy consumption with real world conditions (Schey et al 2012, FleetCarma 2019).There are advantages to the empirical studies, particularly the Charge the North study by FleetCarma (2019) which emphasized the seasonal variability of EV energy consumption in Canadian climates.The Charge the North study found EV charging demand was highest in winter months in Canada, even though distance driven by EVs across Canada was lowest in winter.However, these approaches lack the ability to isolate and quantify the effects of temperature alone on BEV charging impacts, which would allow for repeatability of the analysis in different regions without conducting large scale surveys of charging behavior.Charge the North was also the only study found which covered Canada or a country with a similar cold climate.This is concerning since temperature-related increases to BEV energy consumption are more profound under cold temperatures than hot temperatures (Lohse-Busch et al 2013, Yuksel and Michalek 2015, American Automobile Association 2019).Finally, with regards to climate change, no studies have been found that consider the impacts of climate change on EV performance and charging, as well as its implications on the electrical grid.
This paper aims to project the impacts of BEV charging on electricity demand (daily electricity demand and daily peak demand) in the Greater Toronto and Hamilton Area (GTHA) in Ontario, Canada, accounting for climate variability and change.The GTHA is an ideal area for this study as it allows for the abovementioned BEV analysis in a cold climate which is often not considered, and has detailed household travel survey data available through the 2016 Transportation Tomorrow Survey (TTS) (Data Management Group-University of Toronto n.d.).Although winters in the GTHA are slightly milder than most other major metropolitan regions in Canada (SI section 1.3), the region is of particular interest as it is the most populous in Canada, home to 7 million people or 20% of Canada's population (Statistics Canada 2017).We use detailed vehicle trip data from the 2016 TTS and calculate energy consumption rates with adjustments based on local hourly temperatures.Increases to both daily electricity demand and peak electricity demand are calculated; these two metrics are commonly evaluated in this type of BEV analysis and both have potential impacts on power generation, transmission, and distribution (Muratori 2018).This is done for both historic temperatures as well as future temperatures under climate change; to our knowledge the latter has not been evaluated in the literature.The difference in results between accounting for and omitting ambient temperature variability highlights the importance of accounting more explicitly for temperature in future BEV analyses.

Methods
We estimate hourly BEV charging demand in the GTHA, then add it onto baseline hourly electricity demand to assess the changes to total daily electricity demand and peak hourly electricity demand.The two market penetration levels considered are 10% and 30% of passenger vehicles as BEVs.Penetration of 10% is feasible by 2030 under the International Energy Agency (2021b) projections of 8%-15% EVs (BEVs + PHEVs) worldwide by 2030.Penetration of 30% is feasible by mid-century; BloombergNEF (2020) projects 31% EVs worldwide by 2040 and the US Energy Information Administration (2019) projects 28% EVs worldwide by 2050.Due to the time required for fleet turnover, even scenarios with 100% EV sales by 2035 do not see EV penetration levels above 30% until beyond 2030 (Milovanoff et al 2020), however higher penetration scenarios would further amplify the results presented in this paper.For comparison, Gai et al (2019) assessed 5% and 30% of passenger vehicles as BEVs in the GTHA. Figure 1 provides an overview of the methods, described in detail sections 2.1-2.3.

Temperature and electricity demand profiles
To estimate BEV charging impacts on an hourly basis, data for ambient temperature and baseline electricity demand (without BEVs) are required.In order to consider seasonal variability at the monthly level, profiles containing representative data for each month of the year were produced.Five ambient temperature profiles were created, each containing 12 daily temperature and electricity demand time series (hourly data for a single representative day for each month).This is summarized in table 1. Due to the relation between climate and electricity demand, it was important that both ambient temperature and electricity demand data are for the same temporal period.The year 2019 was used for both data types, with the exception of the climate change profiles.A sensitivity analysis was performed to see the inter-annual variability in BEV electricity demand between 2017 and 2019, showing that inter-annual differences are most substantial in the winter (mild winter vs. cold winter), and when looking at extreme days instead of average days.See SI 1.1 for the full results.

Ambient temperature profiles
Hourly ambient temperature observations for the year 2019 were taken from Toronto Pearson International Airport (Environment and Climate Change Canada 2021b) and used to generate the temperature profiles described below.Data from this station was applied to vehicles across the GTHA; SI 1.2 explores the difference between using a single station instead of multiple, and demonstrates that using a single station is a reasonable approach.The temperature profiles each contain 12 sets of 24 values, 1 set for each month, and 1 value for each hour.The first is the 'monthly average' profile, where the mean hourly temperature across all days of the month is calculated for each hour from 1 to 24; this is repeated for all 12 months.The second and third are the 'hottest day' and 'coldest day' profiles, which contain the hourly temperatures of the day with the hottest/coldest daily mean temperature in each month.
The fourth profile is the 'climate change average' profile.This temperature data came from a variable resolution version of the Community Earth System Model (CESM), also known as VR-CESM (see SI 2.1).This model allows for a much higher spatial resolution (7-10 km grid cells) in the region of focus, in this case the GTHA, compared to the parent simulation CESM (∼100 km grid cells).The model output provides 20 representative samples of the year 2000 under historical forcing conditions and of the year 2040 under conditions when climate change is forced under the RCP 8.5 pathway as defined by the Coupled Model Intercomparison Project, phase 5 (which provided modelling input to the fifth assessment report of the Intergovernmental Panel on Climate Change); see SI 2.1 for further details of the model.Although RCP 8.5 is considered a pessimistic climate change scenario, it is selected to better distinguish signal from noise within the climate projections; further, by 2040, RCP 8.5 does not yet diverge substantially from lower RCPs.The climate change profile for 2040 is compared against the BEV results using the representative samples of the year 2000 (the midpoint of recent 20 year Climate Normals) produced by the same climate model to provide a fairer comparison.All projections are for Toronto Pearson International Airport, which is the same station as the three profiles using 2019 observations.The 'climate change average' temperature profile was calculated the same way as the 'monthly average' temperature profile, except this profile was the mean of the 20 year model sample.The fifth is the 'climate change extremes' profile.This profile consists of representative temperatures for the coldest and hottest day of the year in the climate projections.Out of the hottest and coldest day in each year of the 20 year model sample (based on daily mean temperature), the 25th, 50th, and 75th percentile values are selected.Thus, this profile represents a range of values for the hottest and coldest day within the climate projections.Like the 'climate change average' profile, the 'climate change extremes' profile is also compared to the 20 samples for the year 2000 from the same climate model.

Electricity demand profiles
Hourly Ontario-wide electricity demand values for 2019 were taken from the IESO's publicly available demand reports (Independent Electricity System Operator 2020b).A 'monthly average' profile was produced using the same method as for ambient temperature (section 2.1.1).'Hottest day' and 'coldest day' profiles were produced taking the hourly electricity demand for the hottest/coldest days of the month.In this way they coincide with the 'hottest day' and 'coldest day' temperature profiles.These monthly temperature extremes do not necessarily occur on the same day as the highest electricity demand.However, for 2019, the extremes for daily mean temperature and the highs for daily electricity demand had coinciding dates in 6 of the 12 months and were one day apart for another 4 of the 12 months.
The 'climate change average' and 'climate change extremes' temperature profiles require a corresponding electricity demand profile that is adjusted for the shift in temperatures.This effectively would be present day electricity demand but adjusted for climate change, while holding other factors constant (e.g.population growth, other socio-economic factors).Hourly electricity demand for the Toronto IESO planning region was taken from a regression model developed for projecting Ontario electricity demand under future temperatures; see SI 2.2 for further details of the model.The output was used to produce 'climate change average' and 'climate change extreme' profiles that correspond to their respective temperature profiles.For the 'climate change extremes' profile, the influence of non-climate variables introduces noise to the regression output such that more extreme temperature days do not uniformly correspond to more extreme  b Ontario-wide and Toronto IESO zone electricity demand includes transmission losses.This is for consistency with the IESO hourly demand data used in the analysis.Source: (Independent Electricity System Operator 2020b).electricity demand profiles.Thus, to simplify the analysis, we produce a single representative 'climate change extremes' demand profile as the average of the simulated demand across all years of the 20 year model sample (instead of matching demand to the 25th, 50th and 75th percentiles); results for the added demand from EVs continue to rely on these percentiles as described below.
The GTHA is an ideal focus area for this study due to the availability of detailed household travel data, the presence of cold winters which pose a concern for BEVs, as well as the size and significance of the area; the GTHA being a conurbation of the Toronto and Hamilton census metropolitan areas, first and ninth largest respectively in Canada's 2016 census (Statistics Canada 2017).As the vehicle data used was limited to the GTHA, it would be beneficial to have electricity demand data for the same region.However, the IESO only offers this data at the Ontario-wide level, or broken down into IESO planning regions.We developed an approach to scale the Ontario-wide electricity demand down to the GTHA, with the assumption that the shape of the hourly electricity demand curve for the GTHA is consistent with the Ontario-wide profile; considering that the GTHA already represents nearly half of the province's electricity demand (see table 2), this is likely a reasonable simplification.Nine electricity distribution companies that service parts of the GTHA were selected as a proxy for GTHA electricity demand (see table 2).The scaling factors for GTHA electricity demand was 0.446.Each of the five Ontario electricity demand profiles were multiplied by the scaling factor to create corresponding GTHA electricity profiles.This enables analysis at the GTHA level.An alternate method was used for the climate change comparison, which involves scaling up electricity demand from the 'Toronto' IESO zone (scaling factor of 1.21 to convert to GTHA demand), rather than scaling down from the provincial level.This alternate method was used only for the climate change profiles since the electricity demand profiles under climate change were produced specifically for the Toronto IESO zone.SI 3 contains a comparison of final results under the two methods.
A limitation to this approach is that the service regions for these companies do not fully line up with the regional boundaries that form the GTHA.Based on 2016 Canadian census data (Statistics Canada 2017), approximately 95% of the GTHA population is covered by these selected companies.The remainder of GTHA residents are serviced by Hydro One, which predominantly services rural customers province-wide.In addition, the selected companies cover a region with a population of 7.30 million, which is slightly larger (5%) than the GTHA population of 6.95 million.Therefore, the use of the nine electricity distribution companies as a proxy for the GTHA is not perfect, but it acts as a reasonable estimate.
Finally, current EVs (BEVs + PHEVs) in the passenger vehicle fleet cannot be easily separated from the IESO electricity demand data.Based on new motor vehicle registrations data from Statistics Canada (2021), EVs accounted for just 1.2% of new motor vehicle registrations in Ontario for 2019, lagging behind the national average of 2.9%.While EV market share of the passenger vehicle fleet is not readily available for Ontario, the nationwide estimate for 2019 is just 0.6% (International Energy Agency 2021b).Therefore, the EV charging demand present in IESO electricity data is assumed to be negligible due to the very low 2019 market share of EVs, both as a percent of sales and of the vehicle fleet.Limiting the minimum and maximum SOC is important for preservation of battery lifespan as well as optimal battery performance (Kostopoulos et al 2020); actual 'usable capacity' limits imposed may vary by BEV manufacturer.Reduction of battery capacity in cold weather is not considered, but the size of the effect is smaller than the already assumed SOC constraints.

Temperature adjusted electric vehicle energy consumption
BEV energy consumption was calculated on a trip level, which was then aggregated to the vehicle level to determine charging demand.We follow largely the same approach for quantifying BEV energy consumption at the trip level as Gai et al (2019), but with updated data.Vehicle trip data is taken from the 2016 TTS (Data Management Group-University of Toronto n.d.).Results of the GTAModel V4.0 (Travel Modelling Group 2021), a travel demand model, were used to assign travel distance and time to each vehicle trip in the TTS dataset based on origin-destination pairs.The model's spatial boundaries were the GTHA, meaning trips from GTHA households that started or ended outside of the GTHA were excluded.This accounted for roughly 5% of trips in the expanded dataset for the GTHA.An equation from Yao et al (2013) is used to estimate BEV energy consumption rate (kWh km −1 ) for each trip based on average speed.See Gai et al (2019) for further details of this approach.After this, a unique ID number was assigned to each vehicle in each household to allow for aggregation of BEV energy consumption at the vehicle level.
Next, temperature adjustments are applied to the BEV trip energy consumption.Each temperature profile is passed through a temperature adjustment equation to produce a new matrix of BEV energy consumption adjustment factors for every hour and every month (i.e. based on the five temperature profiles from section 2.1).The equation to adjust BEV energy consumption based on temperature is from Wu et al (2019) (as presented in Miller et al (2020)), which to our knowledge is the most robust model of its kind at present.The temperature adjustment matrix is applied to the BEV trip dataset through multiplying trip energy consumption by the adjustment factor for the hour in which the trip starts.This process is repeated for all months within all temperature profiles.At this point, the trip data was aggregated to the vehicle level to determine charging demand for each vehicle.The travel times of all trips within the day were also combined to determine the vehicle's charging availability (i.e. when and where the vehicle is parked).
We assume infinite battery capacity, which also assumed by Gai et al (2019).This is justified by considering that under the BEV market shares of 10% and 30%, drivers who could not complete their typical travel on a single BEV charge would not adopt a BEV.This assumption is tested in table 3 (section 3.1) where battery capacity limits are applied.Battery capacity is also temperature sensitive, particularly in cold weather; Meyer et al (2012) found that BEV battery discharge capacity decreases by 4%-9% in sub-zero temperatures (−7 • C to −20 • C) relative to 20 • C. The temperature sensitivity of BEV discharge capacity is not taken into account in the present paper since the infinite capacity assumption renders it irrelevant.

Electric vehicle charging demand profiles 2.3.1. Charging cases and assumptions
To determine the timing of BEV electricity demand, assumptions must be made about the charging behavior of the vehicles.Charging rate and charging efficiency are held constant for all cases; all charging is Level 2 at an assumed constant rate of 7 kW, while charging efficiency for Level 2 charging is 0.894 (Sears et al 2014).SI 4 tests this assumption with cases involving Level 1 charging in households where possible; it was found 38% of all charging demand in the GTHA can be completed by Level 1 chargers in residential settings.This potential drops when taking temperature into account, with only 21% of charging being supported by Level 1 charging on an average January day.It is also assumed in all cases that charging occurs exactly once daily, where in reality drivers may charge more than once per day (perhaps at different locations), or less than once per day if their daily driving pattern moderately depletes the battery state of charge (SOC).We employ the following idealized charging cases: 1. Residential Basic Charging.Vehicles charge exclusively at home, beginning after the last trip of the day.2. Residential Time-of-use (TOU) Charging.Vehicles charge exclusively at home, but owners adapt their charging behavior to Ontario's TOU pricing, in which 7pm-7am on weekdays are 'off-peak' hours (Ontario Energy Board 2021).This means if the last trip of the day ends before 7pm, charging does not begin until 7pm.If the last trip ends after 7pm, charging begins immediately after the trip.3. Residential TOU Random Charging.Same as the 'Residential TOU Charging' , but rather than starting charging at 7pm (for vehicles arriving home before 7pm), charging randomly starts between 7pm and 12am, on the hour.This reduces the spike in demand anticipated for when 'off-peak' hours start, and simulates a world in which EV owners wait for off-peak hours to charge but do not cluster at the start of the off-peak period.4. Public Charging.Vehicles charge at an eligible destination (work, shopping, school), during the period in which the vehicle is parked for the longest time.If charging cannot be completed at the eligible destination (or none exist), then the vehicle follows the 'Residential Basic Charging' case. 5. Fully Controlled Charging: Vehicle charging is controlled with the intent of avoiding increases to peak demand.This is roughly estimated by determining the maximum amount of electricity in a 24 h period that can be added to baseline demand without increasing the existing peak, and then comparing it to the total daily charging demand.Thus, this ignores charging availability constraints and is a best case scenario.
Cases 1, 2, 4 and 5 are adapted from cases used by Gai et al (2019) for BEVs in the GTHA, with some differences in assumptions (see SI 5.1 for details).The charging strategies are applied to produce a set of 24 new data fields for each vehicle, containing the charging completed (kWh) in each hour.The temporal resolution of the results are limited to an hourly basis in order to enable direct comparison with the electricity demand profiles.This relies on the assumption that 1 kWh of electricity consumption = 1 kW of instantaneous electricity demand during that hour, which is true if the charging occurs for the entire hour, but false if the charging occurs within only part of the hour.The random element in Case 3 means that the results technically vary between iterations, even though the random variable is uniformly distributed across over 100 000 vehicles in the unexpanded TTS data.An analysis of 100 iterations of this charging case is in SI 5.2; the variance between iterations is largely inconsequential to the results especially once scaled down to a 10% or 30% BEV market share and added on top of baseline hourly electricity demand.
Case 5 is not used in the main analysis, but rather a separate experiment of its own to act as a bounding analysis for the most optimistic possible case.In addition to testing 10% and 30% BEV market shares, we also estimate the maximum amount of BEVs that can be supported under this charging case without a noticeable increase to peak demand.The intent is to scope out the degree to which BEV sensitivity to ambient temperatures reduce the effectiveness of a theoretical fully controlled BEV charging system.

Calculation of hourly electricity demand with BEVs
Each vehicle's hourly charging profile is validated by comparing the sum of the hourly charging to the daily charging demand.This identified a select few cases, or 'error vehicles' where a vehicle cannot meet its charging requirements under the constraints of the charging case and travel patterns.Typically these were vehicles with high daily energy consumption (kWh).Across all scenarios, the average percentage of 'error vehicles' was 0.24% of the unexpanded dataset, and the maximum was 1.2%.Given the assumed BEV market penetrations of 10% and 30%, it was most practical to assume the owners of this small subset of 'error vehicles' would simply not adopt BEVs in the first place.Thus, these 'error vehicles' were excluded from the dataset.This leads to small discrepancies for total BEV electricity demand in the GTHA, an average of 1.3% and maximum of 5.7%, which was corrected for when scaling down to 10% or 30% BEVs.
At this point, the hourly BEV charging demand across the GTHA is calculated by summing up the hourly totals of each vehicle.This process is completed for all charging cases across all monthly temperature profiles.To reach the 10% and 30% market penetrations, the simulation is first run assuming all vehicles (excluding error vehicles) are electrified, and then scaling down accordingly-multiplying the total BEV charging demand by a factor of 0.1 and 0.3.A correction was applied to account for the missing energy consumption from the 'error vehicles' in each scenario to ensure the sample was as close to 10% and 30% of vehicles as possible.This method assumes a uniform distribution of BEV adoption in the GTHA across geographic regions and owner demographics, which is a simplification of what would be expected in reality.
There are 24 sets of BEV hourly charging demand profiles: 5 temperature profiles plus 1 without temperature adjustments, multiplied by the first 4 charging cases.As mentioned, Case 5 is only used as a separate bounding analysis (see section 3.4 for results).Each profile of the hourly BEV charging demand is added onto the estimated GTHA electricity demand.Changes to total daily electricity demand and daily peak electricity demand are calculated for all assessed scenarios.

BEV charging demand
The daily BEV charging demand in the GTHA, without temperature adjustments, was estimated at 2.1 GWh with 10% BEVs, and 6.3 GWh with 30% BEVs.Gai et al (2019) found a daily BEV charging demand of 5.35 GWh with 30% BEVs in the GTHA before charging and transmission losses.When factoring in charging efficiency (Sears et al 2014) to make it directly comparable, their value is 6.0 GWh.Our value is roughly 5% larger even though we follow a highly similar methodology; this difference is reasonable given use of the 2016 TTS in the present paper in contrast to the 2011 TTS in Gai et al (2019).When considering there are 2.67 million vehicles in the expanded dataset, this equates to an average daily charging demand of 7.8 kWh per BEV.
Given the implicit assumption of infinite battery capacity in the study methodology, it is important to test this assumption by estimating the proportion of vehicles that can complete their daily trip pattern with a finite charging capacity and full SOC while charging once daily.Table 3 shows the results of this test using four different battery capacities, and shows how ambient temperatures influence these results through increased energy consumption rates.The effect of temperature on battery capacity itself is not considered in table 3, but the size of the effect (4%-9% decrease in sub-zero temperatures according to Meyer et al (2012)) is much smaller than the assumed SOC constraints of minimum 20% and maximum 90%.This is not to be confused with the error trips as described in the methods, in which charging case constraints were the limiting factor and not battery capacity.Even in the worst case scenario, 90% of drivers can replace their trips with a BEV, which is much higher than the 10% and 30% BEV market shares being tested in the present study.Also recall that in reality the driver would not be constrained to charging exactly once per day which, in addition to the SOC constraints suggest the results in table 3 are conservative.
When adjusting BEV charging demand for temperature, increases of varying magnitudes are observed.Figure 2 shows the percent increase in BEV daily charging demand when adjusting for ambient temperatures, relative to charging demand under standard conditions (i.e.mild temperatures).The percent increases are relative to the daily charging demand without temperature adjustments.The 'monthly average' temperature profile increases daily charging demand by 0%-52%, with an average of 21%.The hottest or coldest day of the month (whichever gives the highest BEV charging demand), increases the charging demand by an average of 37%.The increase is 11% in July's hottest day, and 82% in January's coldest day.While the increase in winter months due to temperature is far higher than the summer months, the hottest summer days are still pertinent as that is when the annual peak power demand typically occurs in Ontario.
In the context of GTHA power demand, without adjusting for temperature, BEV charging accounts for 3.3%-4.1% of daily GTHA demand with 30% BEVs.With the 'monthly average' temperature profile, this range is 3.4%-5.0%,and for the 'hottest day' and 'coldest day' profiles, the same range is 3.3%-5.5%.The range exists due to the range of baseline GTHA power demand values accounting for seasonal differences.The proportional demand of BEVs under the 'hottest day' and 'coldest day' profiles is not much larger than under the 'monthly average' profile.This is due to the baseline demand also increasing on the 'hottest day' and 'coldest day' profiles compared to the 'monthly average' profile, diminishing the change attributed to BEVs.

BEV contributions to peak electricity demand
While the electricity consumed by BEVs in this analysis is determined by travel patterns and ambient temperature, the timing for the electricity demand on the grid is determined by the charging behavior.This hourly electricity demand, when added on top of the baseline hourly electricity demand, has potential to raise the magnitude and/or shift the timing of the peak demand.Ontario typically sees two peaks within the year; the annual peak typically in the summer, and a smaller but distinguishable peak in the winter.SI 6.1 contains baseline electricity profiles for a select few months.This section focuses specifically on results for January and July, as they are the months in which the winter and summer peaks occurred for 2019.SI 6 contains tables of the full results across all scenarios and seasons.
Figure 3 shows the hourly distribution for all four charging cases (without temperature adjustments) before being superimposed on GTHA hourly electricity demand.Note that introducing temperature adjustments both raises and widens the peaks since the effect is to lengthen charging time for each vehicle, creating more overlap between vehicles; thus, the hourly charging demand scenarios do not simply scale uniformly with temperature.Nevertheless, figure 3 provides a reasonable visual approximation to the shape  of the load under a wide range of temperatures.The figure shows 'Residential Basic' and 'Public Basic' both peak at about 15% of daily charging demand, with the former peaking between 6 to 7pm and the latter peaking between 8 to 9 pm.The 'Residential TOU' charging appears problematic, with 46% of daily charging completed between 7 to 8pm.This case illustrates the 'rebound peak' , which is a spike in demand produced as a response to changing electricity rates under TOU pricing (Muratori and Rizzoni 2016).It serves as more of a bounding scenario than a probable case, since full adherence to the assumptions of the charging case is highly unlikely.Finally, the 'Residential TOU Random' charging is effective at smoothing demand and shifting it further off-peak, with a peak daily charging share of 17% between 9 to 10 pm.The effects of charging timing on BEV's contribution to peak demand is not novel (Hanemann et al 2017); however, it is shown here to provide context to the wide range in results for BEV's contribution to peak demand specifically in the GTHA under these cases, and the results that follow.
The power demand profiles created for the GTHA contain the baseline peak hourly loads.In the monthly average demand profile, daily peak demand without added BEVs was between 6.9 and 9.0 GW.The units here are technically GWh h −1 due to the hourly resolution of all data.Using the hottest and coldest days, peak demand becomes 9.4 GW in January and 9.7 GW in July.The peak power demand does not necessarily have to coincide with the temperature extremes, but in this case the July's hottest day value was within 100 MW (roughly 1%) of the value for the true 2019 peak hourly demand, which occurred on a different day in July. Figure 4 shows the increase in peak demand due to BEV charging across the year, with error bars representing different charging profiles.The values are for the hottest or coldest day of each month, whichever produced the highest peak demand.The increases ranged from 1.0%-12% for 10% BEVs, and 2.9%-29% for 30% BEVs across all charging cases; the annual averages ranged from 2.4 to 5.8% for 10% BEVs, and 11%-20% for 30% BEVs.Thus, the contribution of BEVs to peak demand is larger percentage wise than the increase in total daily charging demand seen in section 3.1 (3.3%-5.5% with 30% BEVs).
Figure 4 also illustrates the importance of focusing on BEV demand in the winter and summer compared to the fall and autumn, as the baseline peak demand is already highest in those months.In fact, for charging cases 1-3, the addition of BEVs was enough in some cases to shift the annual peak from the summer to winter.The annual peak without BEVs was 9.7 GW in July's hottest day, with January's coldest day having a peak of 9.4 GW.Adding 30% BEVs changed July's peak to a range of 9.8-10.2GW, while January had a range of 9.5-10.5GW.Thus, the peak shifted from July to January for 3 of the 4 charging cases.
Another observation in figure 4, but even more clearly visible in figure 5, is that the contribution of BEVs to peak demand in absolute values is larger for January than July.This is consistent with daily BEV charging demand, where winter also sees the largest gain as shown in figure 2. The most important feature in figure 5 is how much the adjustment of BEV charging demand for ambient temperatures increases peak demand (dark part of the bars).As seen in figure 5, the adjustment of BEV energy consumption for temperature increases peak demand by roughly 330-700 MW (depending on the charging scenario) on January's coldest day and 35-130 MW on July's hottest day, assuming 30% BEVs.When comparing these temperature-related increases to baseline peak demand without BEVs, the increase in peak demand is an additional 0.3%-0.9% in July, and an additional 3.2%-4.7% in January.While this effect is much more profound for January, it is still important to consider in July since the summer is when the annual peak power demand typically occurs.

Climate change
First, the difference in BEV charging demand caused by climate change by 2040 is evaluated using model output from VR-CESM under RCP 8.5. Figure 6 illustrates the BEV daily charging demand under a monthly  average profile for a typical historic year (2000) and the future (2040) under climate change.There is a 5.8% decrease in January and a 3.3% increase in July due to climate change.On average across all months, there is a net decrease of 2.6%.The general trend seen in figure 6 is of decreases to BEV energy consumption in the winter offset by increases in the summer.This is in line with the baseline projected power demand under climate change provided by the regression model.For the monthly average profile of daily baseline electricity demand (excluding BEVs), daily demand decreased 2.7% in January, and increased 8.3% in July between the present and future climates.
The estimated increases to peak demand from BEVs also changed in the climate change extremes profile when comparing present day to the future under RCP 8.5.Table 4 shows the baseline peak electricity demand in the GTHA on extreme days for present and future climates, as well as the peak when including 30% BEVs across charging scenarios 1-4.While baseline peak demand changes between the present and future by −3.3% and +9.1% on the median coldest and hottest day of the year respectively, the relative contribution of BEVs to peak demand does not change substantially.As seen in table 4, 30% BEVs on the median coldest day of the year increases GTHA peak demand by 1.2-3.6GW (15%-43%) in the year 2000, and 1.2-3.5 GW (15%-44%) in 2040 under climate change.On the median hottest day of the year, these values are 0.3-2.7 GW (3.4%-31%) in the year 2000 and 0.3-2.7 GW(3.4%-28%) in 2040.
The temperature variability of the hottest and coldest days of the year also had a small but nontrivial effect on the contribution of BEVs to GTHA peak demand, and this holds true in the future under climate change.This can be seen in table 5, where the change in peak demand between the 25th and 75th percentile of daily mean temperature for extreme day of the year is on the order of 120 MW or less.This is solely for the residential basic charging scenario.

Benefits of controlled charging
The adjustment of BEV energy consumption based on ambient temperature did reduce the maximum amount of BEVs that could be supported with a controlled charging scenario that avoids adding to peak demand.Only January's coldest day and July's hottest day were considered in this section since the extreme temperature days are most pertinent for this type of analysis.On January's coldest day, the baseline demand curve allows for a hypothetical 100% BEV market share without increasing the peak using an idealized controlled charging scenario.When adjusting for temperature, the highest possible market share with the given constraints drops to 71%.In July's hottest day, a market share of 100% BEVs can be sustained without adjusting for temperature.This value does not drop when adjusting for temperature meaning a 100% BEV market share can still be supported.Recall the key assumption in this charging case in which charging availability is ignored; this means the results of this section are upper bounds rather than projections of the most likely outcome for an optimized charging scenario.A table of the full results for this analysis is in SI 6.
The amount of BEVs that can be supported under fully controlled charging without increasing the peak demand depends on other factors in addition to the BEV charging demand.For example, July has a larger 'valley' in the early morning than January; in July the minimum demand is roughly 3000 MW lower than the peak, compared to January where the minimum is roughly 2300 MW lower than the peak.Generally speaking, the larger the 'valley' seen in the early hours of the morning, the more BEV charging that can be supported without affecting peak demand.Thus, the ability to support 100% BEVs under fully controlled charging on July's hottest day is not only due to the lower temperature-adjusted BEV charging demand (compared to January), but the characteristics of the baseline power demand curve for the GTHA.

Discussion
The increase in charging demand associated with temperature extremes does reduce the amount of GTHA drivers that could complete their trips using current BEVs.However, even for the worst-case weather conditions (January's coldest day), BEVs with small 40 kWh battery packs can meet daily travel needs for 90% of GTHA drivers.This number increases to 99% with a 100 kWh battery pack, which is at the high end for current BEV models.These results are not surprising given Needell et al (2016) found that a 2013 Nissan Leaf (24 kWh capacity) could meet travel needs of 87% of vehicle-days in the US; they also accounted for the temperature dependent auxiliary energy consumption.The fact that BEVs can meet an even greater majority of GTHA drivers' needs on an average weekday even under severe conditions is an important point, given that range anxiety is a well-known barrier to widespread BEV adoption (Li et al 2017).
Even when looking at BEVs in the GTHA alone, the impact of BEV charging on peak demand is substantial enough to be noticed at the Ontario-wide level.This is exacerbated by the increases in BEV charging demand due to ambient temperature.The increase associated with BEV temperature effects alone, as seen in figure 5, equates to between 40 and 130 MW in July's hottest day, and 330-700 MW in January's coldest day (at 30% BEV penetration).This number does not include transmission losses, and would of course be even higher if BEVs were considered across all Ontario or higher rates of BEV penetration.
While the present work assessed temperature adjusted BEV energy consumption across all seasons, emphasis was placed on the impacts of BEVs to the grid during the hottest and coldest extremes of the year.The dual peaks of summer and winter are of importance to the Ontario electricity grid and both seasonal peaks are considered in official projections by stakeholders (Independent Electricity System Operator 2020a).
A moderate BEV market share (30%) was enough unilaterally to shift the annual peak in 2019 from July to January in 3 of 4 charging cases.However 2019 was a colder than normal year and the conclusion may not extend to other years.In Ontario's case, this shift does not necessarily pose a supply issue as there is currently a surplus of installed capacity and effective generation capacity is somewhat higher in the winter than the summer (Independent Electricity System Operator 2020a).Nevertheless, future increases in demand in Ontario are coupled with decreases in capacity as current plants reach their end of life and current contracts expire (Independent Electricity System Operator 2020a), necessitating careful consideration of future contributions from seasonal and peak EV load for accurate power reliability planning.This shift could potentially be exacerbated by the general phasing out of fossil fuels in favor of electricity, for example with electrification of household heating.
Even if the supply system is capable of meeting the projected peak demand increases due to temperature-adjusted BEV charging demand, reducing peak demand is still important for environmental reasons.Considering the 2017 marginal emission factors for Ontario developed by Gai et al (2019), the peak demand periods tend to have highest reliance on natural gas and thus the highest marginal emissions.The higher the peak demand, the larger the reliance on natural gas for marginal generation, and thus the larger the effective emission of BEVs.
The influence of climate change on BEV charging demand and subsequent related impacts has also been analyzed here.Results suggest that the proportional impact of BEV charging on overall electricity demand does not substantially change due to climate change.This is because overall electricity demand is also climate sensitive and shifts in the same direction as BEVs do in response to changes in temperature.On an absolute scale, however, the additional contribution of BEVs to peak demand may be especially problematic during extreme temperatures since the larger BEV demand is being added to an already elevated baseline peak demand.Future projections of BEV impacts should still take climate change into account in order to best represent future conditions in which mass adoption of BEVs would occur.
The maximum number of BEVs that can be supported on a fully controlled charging network without noticeably adding to peak demand is reduced after taking temperature into account.On January's coldest day, the maximum BEV share supported falls from 100% to 71%, while on July's hottest day it is 100% in both cases.Practically speaking, the reduction to a maximum 71% BEVs on January's coldest day would in fact be a limiting factor for the entire year.Nonetheless, filling of the valley in GTHA electricity demand would still carry benefits.First, the filling of the early morning valley offers reduced GHG emissions from the associated electricity generation, seen by other studies in the GTHA (Gai et al 2019, Tu et al 2020).Second, the valley filling may benefit the electricity supply system.Foley et al (2013) found BEV charging during off-peak hours in Ireland allows for better utilizations of base-load plants as well as wind power, leading to an improvement in system efficiency.The charging case in the present study provides only an approximate solution and neglects charging availability constraints.However, the exercise was intended to show the difference temperature makes in the capabilities of fully controlled charging, which it achieved.

Conclusions
This study estimated the impacts of BEV charging on the electric grid in the GTHA, specifically quantifying the impact of additional BEV charging demand attributed to climate variability and potential future climate change.The use of the TTS 2016 data allowed for a more detailed approach than what is often seen in the literature, calculating BEV energy consumption at the trip level and adjusting it based on the hourly temperatures at which the trip occurred.The study focuses on a region with cold winters, and regions with such climate are sparsely covered in the literature.The GTHA in particular was an ideal study area within Canada due to the large population coverage of 7 million or approximately 20% of Canada's population (Statistics Canada 2017), and the presence of high quality travel demand study data through the TTS.Nevertheless, we caution that climate and travel patterns vary across Canada (and other countries with cold climates), which will influence quantitative results, but likely not the qualitative conclusions.
Within the GTHA, ambient temperature has a strong influence on BEV daily charging demand.Adjusting for ambient temperatures led to charging demand increases of 0%-52% under the monthly average profile, and 5%-82% under the extreme days of each month.For peak demand, taking temperature into account for 30% BEVs led to an additional increase in peak electricity demand of 3.2%-4.7%on January's coldest day, and 0.3%-0.9% on July's hottest day.Winter saw the highest increases across both metrics while summer saw the lowest increases.Demand in the summer months is still important however since it is when electricity demand is typically highest.Climate change's impact on BEV charging is small in absolute terms compared to its impact on broader electricity demand, but may still be of concern when added to an already elevated baseline electricity demand and should still be considered for accurate demand projection.Temperature-adjusted BEV charging demand did lower the maximum market share sustainable of an approximated optimized charging case from 100% to 71%, however 71% is still much higher than the projected values for BEV market share in the near future.
This work relies on several assumptions which could be addressed with future work and more detailed information.Charging efficiency was held constant when in reality it may vary due to ambient conditions, although more research is needed to quantify that effect (Sears et al 2014).Additionally, the charging cases were all variations of charging exactly once per day, with all vehicles adhering to the same idealized charging strategy.Future work can explore a wider variety of charging patterns, including a heterogenous mix of strategies, in addition to allowing multiple charging events per day or less than one charging event per day.Resolution of the analysis could be further improved by simulating a heterogenous mix of vehicle models and efficiencies, however we speculate that aggregate charging demand would roughly converge to the profile of the average representative vehicle used in this study.Another variable held constant was vehicle km travelled (VKT).The TTS 2016 data was collected in the fall and does not capture seasonal driving patterns.In reality, VKT fluctuates seasonally.The Charge the North survey observed that in Canada, monthly BEV charging demand is highest in winter, even though VKT was typically lower in winter than summer (FleetCarma 2019).This suggests that the higher BEV energy consumption rate seen in winter could be offset by reduced VKT, and vice versa in the summer.There are also a multitude of technological innovations and behavioral changes possible within the near future, i.e. by the time the BEV market grows to the shares explored in this paper, however they were not explored in the present work.This includes, but is not limited to, BEV performance improvements, charging infrastructure accessibility and technology improvements, travel pattern changes, and decarbonization of other sectors through electrification.Finally, PHEVs were not considered since they have a lower market share than BEVs (International Energy Agency 2021b), but can be considered if appropriate assumptions are made.PHEVs add an extra layer of complexity as the proportion of VKT completed on electricity is a variable that introduces many new factors to consider.
Despite the above limitations, the results show convincingly the need to for planners, policy makers, and utility companies to take temperature variability into account for accurate projections of BEV impacts on the electric grid, particularly in colder climates.Smart charging technology and similar emerging technologies to reduce BEV impacts on the grid will likely see reduced effectiveness in climates with high temperature variability; this should be taken into account by stakeholders involved in the development and implementation of these technologies.Finally, there is an opportunity to extend this type of analysis to regions with diverse geographies and climates, as well as take temperature into account for other electric grid impacts not considered here such as voltage stability and power losses.

a
Delivered electricity excludes transmission losses.Source: (Ontario Energy Board 2020).

Figure 2 .
Figure 2. Percent increase in BEV daily charging demand when adjusting for ambient temperature, compared to when not adjusting for it.

Figure 3 .
Figure3.Hourly distribution of daily BEV charging demand for each charging case (without temperature adjustment).Fully controlled charging (Case 5) is not shown as its hourly distribution varies dynamically with the non-EV load, depending on the profile to which it is applied.

Figure 4 .
Figure 4. Average increase in peak daily demand due to BEVs for hottest or coldest day of each month, whichever produced the highest peak demand.The hottest day produced the highest peak demand from June to October, while the coldest day produced the highest for the remaining months.The blue bar represents the average increase across charging cases 1-4.Error bars represent the minimum and maximum values based on these charging cases.

Figure 5 .
Figure 5. Increase in peak electricity demand with 30% BEVs in the GTHA for key months, shown in both January and July for each of 4 charging scenarios.The lower value (lighter shade) is from 'Ignoring Temperature' while the upper value (darker shade) is from the 'Hottest Day' or 'Coldest Day' temperature profiles.

Figure 6 .
Figure 6.Comparison of daily BEV charging demand, with 30% BEVs, between a typical historic year (2000) and the future (2040) under climate change.Both present and future temperatures are an average of 20 model years from VR-CESM representing year 2000 and year 2040 conditions.

Table 1 .
Summary of ambient temperature and electricity demand profiles.

Table 2 .
Subset of Ontario electricity distributors selected to represent the GTHA.

Table 3 .
Percent of BEVs in GTHA that can complete daily travel requirements under different temperature conditions and with different battery capacities.Assumed battery capacities do not necessarily correspond to specific vehicles but provide a reasonable range of values based on the 2021 model year.Examples at the higher and lower ends respectively are the Tesla Model S Long Range (100 kWh capacity, 663 km range) and the Nissan Leaf SV (40 kWh capacity, 240 km range) (PlugNDrive 2021).b Assumed usable capacity was calculated with an assumption of maximum SOC being 90% and minimum SOC being 20%. a

Table 4 .
Peak electricity demand (MW) in the GTHA with and without BEVs on extreme temperature days, across charging scenarios 1-4.The median of the daily mean temperature across all 20 years is taken from the subset of the hottest and coldest days of the year.Average baseline peak demand is calculated by taking the mean of the extreme hot or cold day of the year across each year of a 20 year model sample. a

Table 5 .
Peak electricity demand (MW) in the GTHA with and without BEVs on extreme temperature days using the residential basic charging case.The 25th and 75th percentile values of the daily mean temperature across each year of a 20 year model sample from VR-CESM, from the subset of the hottest and coldest days of the year.Average baseline peak demand is calculated by taking the mean of the extreme hot or cold day in the year across each year of a 20 year model sample. a