As of December 2020, U.S. consumers can choose from more than 50 light-duty EV models that span multiple vehicle classes, markets, and a wide MSRP price range from $27,500 to more than $100,000. Projections for continued EV growth through the present “second decade” of mass deployment are varied, but many suggest a sustained exponential growth as evidenced by Figure 1-1, which shows future market share as a fraction of new vehicle sales. EVs are increasingly seen as a win-win solution by many policymakers, in that they can provide benefits to consumers, automakers, and utilities, while also reducing environmental impacts associated with tailpipe emissions. Furthermore, while much public attention is focused on Light Duty vehicle markets, additional opportunities exist in Medium Duty and certain Heavy Duty applications. Despite substantial progress, widespread optimism, and proactive policy support, new and nontrivial barriers remain. These barriers may simultaneously threaten both broader adoption and certain beneficial outcomes of EV growth. Among the most critical and poorly understood, is the need to ensure environmental benefits live up to their promise, in particular under deep deployment scenarios where EVs comprise more than 10% of the future fleet by 2030, bud curing and demand commensurate new supplies of electric power in both time and space. It is desirable to assess the criticality of environmental impacts, so as to quantify the levels of decarbonization enabled by EV penetration.
However, empirical methods are presently insufficient for near term projections, due to uncertainties related to related to charging levels, charging times and the spatial temporal impact of different electricity generation mixes. Further compounding this challenge are insufficient data on EV growth and uncertain adoption rates. What happens, for instance, when EV load growth will require 20% more power demand than is currently forecasted in existing integrated resource plans, which already must also provision for an approximate 15% peak reserve margin? What happens when this average increase in power demand is considered on an hourly or seasonal basis, spiking to much greater shares of reserve . Assuming that utilities embrace an opportunity to sell more kWh to meet new market demands, what assurances are in place to protect the environmental footprint of new load growth?This research requires the synthesis of three independent models developed uniquely by the research team in the areas of vehicle propulsion to satisfy prescribed trip/travel demands for a range of vehicle technologies, EV charging profiles to reflect typical approaches for light duty vehicle use cases, and grid generation dispatch with commensurate consideration of emissions intensities for CO2 and major criteria pollutants. The team has an established track record of developing high-fidelity sub-system models and applying them to both generalizable and regional scenarios.
The team has leveraged more than three years of prior efforts, during which time we acquired and conditioned open-source data and amassed specifications for five representative alternative vehicle architectures, customized datasets for regional electric power dispatch , and numerous travel route pathways. The scope of this project is to update and develop new, more accurate sub-system models and datasets that are relevant, representative, and granular. As described in the original proposal, the team has leveraged these data and iterated upon prior sub-system models with the express purpose of devoting focused attention to integration, simulation, and assessment of results and implications. The end result, therefore, is an integrated model that pulls high-fidelity data from real-world use cases to generate a range of simulations. The simulations will be primarily used to draw comparisons, understand the impact of fundamental assumptions around charging behavior and grid emissions, and develop initial guidance around the relative merits of EVs under representative use cases.The first step in the analysis is the refinement of physics-based vehicle energy consumption models that permit comparison of a range of vehicle architectures that utilize energy from disparate primary sources . A parallel task is to impose upon the vehicle propulsion model a range of driving cycles that can best approximate typical characteristics of representative use cases. Our methodology affords access to established data and extends prior vehicle propulsion energy and emissions analyses.
As a parallel input, the team has utilized individual EPA dynamometer schedules, replicated the 5-cycle fuel economy label weighting protocol, and also consulted independently derived travel demands from representative use cases . A detailed discussion of the theory, model development, source data, and initial applications can be found in [13]. Minoradjustments have been made to vehicle modeling to accommodate key vehicle classifications of interest , and to ensure appropriate reasoning to walk from prescribed EPA dyno schedules to the 5 cycle weighted means, and further to practical interpretations of household travel for representative use cases.As discussed above, we adapt the physics-based power train models developed in [13] to accommodate target vehicle technologies of interest. This includes baseline vehicles , as well as electrified power trains . For the purposes of this study, only pure battery electric vehicles have been evaluated. However, all relevant LDV vehicle technology models have been developed and coded, meaning PHEV analysis is readily available and may be of interest. Owing to their unique architecture which operate as both HEVs and EVs, depending on the battery state of charge, the environmental impacts of PHEVs can be estimated as a weighted mix of the individual impacts of HEVs and EVs respectively. To facilitate a direct comparison among vehicles using dissimilar energy sources, we identify vehicle specifications for a given light-duty vehicle classification and hold constant key parameters such as vehicle power output, vehicle footprint, passenger and cargo capacity, and so forth. Table 2-1 below depicts some of these operative specs.
Note that some differences are inherent in other categories, such as vehicle curb weight. But these have been left as specified by the OEM, under the argument that mass-production specs are reflective of the current state of the art and therefore an excellent proxy for the inherent tradeoffs or interactions to deliver vehicles of similar performance.The five distinct driving cycles that comprise the EPA test and labeling protocol are well documented and widely used for comparative analyses. The three 23°C tests include a derivative of the Urban Dynamometer Driving Schedule known as the Federal Test Protocol , the high-acceleration aggressive driving schedule identified as the Supplemental FTP , and the Highway Fuel Economy Driving Schedule . The 35°C drive cycle is the Air Conditioning Supplemental FTP driving schedule referred to as SC03. The -7°C cold weather test schedule repeats the original FTP at the reduced temperature. As mentioned, curing weed the study has adopted the EPA “5-cycle” protocol and created an approach whereby a weighted mix of driving schedules is obtained to approximate major modes . Please see Appendix A for more details about the weighting of the constituent driving cycles, and the governing formulae. With the original development of the vehicle architecture models, and assumptions around the weighted driving cycle protocols, the team’s next step was to develop a MATLAB/Simulink code that generated a series of energy consumption values based on inputs of vehicle type and driving cycle. These intermediate outputs were then combined to generate effective fuel economy values, analogous to the EPA 5-cycle approach, for the stipulated categories . This was done and a set of energy consumption outputs were generated. These outputs are depicted in Table 2-2.Regarding EV charging behavior, we consider about four primary sources of data to establish representative EV charging profiles. Two are explicitly for residential charging, one is explicitly for workplace charging, and the fourth speaks with survey data collected for both and other categories . The authors acknowledge that there is a growing body of literature on the subject of charging behavior by numerous transportation research centers of note . The authors further suggest that the approach taken herein is appropriate for the purposes of these comparative analyses. It is of note that this research study draws from a combination of analytical and empirical sources of information and data to develop its charging profiles and use cases. Included in this, as detailed below, are first hand studies by researchers involved in the study, utility rate structures that are specific to EV users in the target region, and real-world observed EV charging behavior for a selected network in downtown Atlanta. None of these is unique, and similar approaches are used elsewhere. This, this approach is intended to demonstrate the types of sources of data that this methodology may leverage, and to showcase how they may be applied in a representative set of simulations and outputs. As a first step, we refer to synthetic data generated by a separate research team from Georgia Tech that is evaluating the benefits and challenges associated with smart charging algorithms. Second, we consult the Georgia Power Electric Vehicle Rate scheme, which provides customers with EVs at a deeply discounted rate during off-peak times. In exchange, the rate is tiered, with a relatively expensive energy rate during summer afternoons, and then a fairly nominal price during all other times of the day/year. Third, we refer to data from a Charge Point dashboard portal and database that has been aggregated for workplace charging on the Georgia Tech campus since about 2015. An example of some of this data for a sample month is presented below. It is noteworthy that typical workplace charging occurs in two waves: morning and immediately following the noon hour. A part of the explanation for this has to do with policy: the GT Parking administrator provides a much lower rate for the first 4 hours and then adjusts this to several fold higher to incentivize the EV owner to vacate the parking space and permit additional EV owners an opportunity to charge. The dashboard data is extremely valuable in providing statistically significant information , that can inform real, not perceived or stated, preferences.The team has developed a system-of-systems model that enables the integration of the three sub-system models described in this section. Doing so enables comprehensive and quantitatives imulations of EV deployment for multiple driving cases, under varying EV charging and grid scenarios. The produced MATLAB/Simulink model consists of two user-loaded look-up tables for the selected grid emissions profile and EV charging profile that are imported using the initialization code. The look-up tables take the form of a time series with 1440 distinct time stamps, equal to the number of minutes in a day. The emissions profiles available to the team consisted of hourly emissions rates. The emissions rate during a given hour was assumed to be the same for each minute in that hour. In this manner, each hour of emissions was dissected into 60 periods to achieve 1440 rows of data. By creating minute-by-minute lookup tables, the model is able to stop accumulating grid emissions the same minute the vehicle’s battery is recharged, minimizing returns of surplus charge. Besides loading look-up tables, the initialization code also provides an opportunity for the user to calibrate the energy target . For this study, the energy targets were calculated for each simulated use case using our Vehicle Energy Model described in the previous section. The initialization code and input parameters utilized in this study can be found in Appendix C. Once the initialization code is run, the Simulink component of the model references the loaded look-up tables and parameters, integrating the sub-system models using a series of logical arguments. The completed simulation provides an aggregated output that describes the cumulative grid emissions attributed to the simulated recharge event. These emissions totals are easily converted to a per-unit distance rate. The architecture of the Simulink model can be seen in Figure 2-9.When comparing resulting emissions rates under each grid assumption, it is immediately clear that accepting a monthly or annual average grid emissions rate fails to capture the significant variance that occurs throughout a given day. At higher temporal resolutions, daily grid emissions profiles begin to emerge that have important implications for finding the true environmental benefits of EVs and how those benefits vary depending on the timing of charging events. On average, CO2 emissions per kilometer for an EV charged under the Residential Overnight or Residential Evening charging profiles were found to be less than an EV charged under the Workplace Morning or Workplace Afternoon profiles, especially in the summer and shoulder months. For example, an EV performing the Short Commute trip and charging with the Residential Overnight charging profile in August was found to emit over 3% less CO2 per kilometer when using hourly grid emissions profiles compared to annual averages and nearly 7% less CO2 per kilometer compared to monthly averages.