Effects of Plug-In Hybrid Electric Vehicles on ... - ACS Publications

Jul 15, 2010 - Aunitcommitmentanddispatchmodelwasusedtoestimate the charging patterns of PHEVs and dispatch power plants to meet electricity demand...
0 downloads 0 Views 921KB Size
Download PDF
Environ. Sci. Technol. 2010, 44, 6256–6262

Effects of Plug-In Hybrid Electric Vehicles on Ozone Concentrations in Colorado G R E G O R Y L . B R I N K M A N , * ,†,‡ PAUL DENHOLM,‡ MICHAEL P. HANNIGAN,† AND JANA B. MILFORD† Department of Mechanical Engineering, University of Colorado, Boulder, Colorado, 80309, and National Renewable Energy Laboratory, Golden, Colorado, 80401

Received April 5, 2010. Revised manuscript received June 21, 2010. Accepted June 30, 2010.

This study explores how ozone concentrations in the Denver, CO area might have been different if plug-in hybrid electric vehicles (PHEVs) had replaced light duty gasoline vehicles in summer 2006. A unit commitment and dispatch model was used to estimate the charging patterns of PHEVs and dispatch power plants to meet electricity demand. Emission changes were estimated basedongasolinedisplacementandtheemissioncharacteristics of the power plants providing additional electricity. The Comprehensive Air Quality Model with extensions (CAMx) was used to simulate the effects of these emissions changes on ozone concentrations. Natural gas units provided most of the electricity used for charging PHEVs in the scenarios considered. With 100% PHEV penetration, nitrogen oxide (NOx) emissions were reduced by 27 tons per day (tpd) from a fleet of 1.7 million vehicles and were increased by 3 tpd from power plants; VOC emissions were reduced by 57 tpd. These emission changes reduced modeled peak 8-h average ozone concentrations by approximately 2-3 ppb on most days. Ozone concentration increases were modeled for small areas near central Denver. Future research is needed to forecast when significant PHEV penetration may occur and to anticipate characteristics of the corresponding power plant and vehicle fleets.

Introduction Plug-in hybrid electric vehicles (PHEVs) have both an internal combustion engine and an electric motor. Onboard batteries store electricity from the electric grid to power the vehicle in addition to the gasoline engine. PHEVs have been promoted for several reasons. The use of electricity could help diversify the transportation fuel supply. Although initial vehicle costs are likely to be higher than for conventional gasoline or hybrid vehicles, fuel costs are significantly lower with electricity compared to gasoline. For the average vehicle in this study, traveling 25 miles would cost approximately $0.70 using electricity and $2.36 using gasoline, assuming electricity is available at $0.11/kWh and gasoline at $2.32 per gallon (1). PHEVs also have the potential to reduce trans* Corresponding author current address: Gregory L. Brinkman, 1617 Cole Blvd., Mailstop 301, Golden, CO, 80401; e-mail: gregory. [email protected]; phone: 303-384-7390. † University of Colorado. ‡ National Renewable Energy Laboratory. 6256

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 16, 2010

portation-related emissions of some pollutants. For example, in the United States, total CO2 emissions per vehicle mile (including emissions from electricity generation) could be reduced by 25-50% with PHEVs compared to conventional hybrid electric vehicles (2). On the other hand, by increasing electricity demand, PHEV use could lead to net increases in emissions of sulfur oxides (if not capped) and other pollutants released from fossil fuel combustion for electricity generation. PHEV use would also shift the temporal and spatial distribution of local air pollutants, such as nitrogen oxides (NOx) and volatile organic compounds (VOCs), from vehicle tailpipes to central electric generating units (EGUs). Ozone is formed in the atmosphere through nonlinear reactions involving NOx and VOCs in the presence of sunlight. Ambient ozone concentrations have been correlated with adverse health effects, including hospital admissions, decreased lung function, and premature mortality (3). The U.S. Environmental Protection Agency (EPA) recently lowered the National Ambient Air Quality Standard (NAAQS) for the 8-h average ozone concentration to 75 parts per billion (ppb) to protect human health (4). Understanding the prospective impact of PHEV penetration on ozone levels is of particular interest, because it requires detailed analysis of associated emissions changes as well as chemistry and transport modeling. Previous studies have estimated a wide range of changes in NOx emissions due to the use of electricity to power vehicles, depending on assumptions about environmental regulations, power plant emission characteristics, and whether emissions during the manufacturing process are included. In a report for the California Energy Commission, TIAX estimated that NOx emission factors of new cars in 2012 could be reduced from 0.32 g/mi to 0.01 g/mi if electricity used to power vehicles comes from natural gas and 20% renewable sources and new EGUs in California are required to offset NOx emissions (5). In contrast, Gaines et al. (6) estimated that in 2015, lifecycle NOx emissions from new PHEVs would be 38% higher than those from new gasoline vehicles, if the electricity came from current coal-fired power plants without a NOx emission offset requirement. Using a unit commitment and dispatch model of existing EGUs, Parks et al. (7) estimated that PHEVs would reduce NOx emissions per mile by 65-70% in Colorado using electricity from current power plants as the primary fuel. Sioshansi and Denholm (8) used a unit commitment and dispatch model to study emissions effects of PHEV penetration in Texas, based on the existing power plant fleet. At low PHEV penetrations (up to 15%), NOx emissions from electricity generation were actually reduced in the model due to load shifting away from relatively inefficient peaking units. Only a few previous studies have gone beyond estimating impacts on emissions to model impacts of PHEVs on air quality. In a relatively comprehensive study, the Electric Power Research Institute (9) used the Community Multiscale Air Quality (CMAQ) model with 36-km horizontal resolution (and without a plume-in-grid treatment of power plant plumes) to simulate the prospective effects of PHEVs on ozone, particulate matter, and other pollutants across the contiguous U.S. The PHEV scenario assumed 40% market penetration of PHEVs in 2030, that electricity powered 20% of all vehicle miles traveled (VMT), and total electricity demand increased by 6%. Compared to a base case that assumed a mixture of conventional gasoline and hybrid vehicles, fourth highest annual 8-h average ozone concentrations were reduced by 1 to 2 parts per billion (ppb) over the eastern U.S. and near large urban areas across the country. 10.1021/es101076c

 2010 American Chemical Society

Published on Web 07/15/2010

Ozone concentration increases were modeled near a few large power plants in the central U.S. Modeled eighth highest annual 24-h average PM2.5 and PM10 concentrations in the eastern U.S. were reduced by up to 1 µg/m3. Changes in PM10 and PM2.5 concentrations in Colorado (and most areas of the western U.S. outside California) were negligible. In a regional study, Thompson et al. (10) used the CAMx air quality model with 12-km horizontal resolution to explore impacts of hypothetical PHEV use on ozone in the Northeast during a severe photochemical event in 2002, assuming that PHEVs would be charged at night using spare coal-fired power plant capacity. NOx emissions from EGUs were assumed to have no cap and trade program limits, and vehicle emission characteristics for the base case were based on EPA data for the 2002 vehicle fleet. For this simple hypothetical scenario and a time period with very high ozone concentrations, they found mixed impacts, with reductions of 2-6 ppb in peak 8-h average ozone concentrations in urban areas, but increases of up to 8 ppb in some relatively small areas. Colella et al. (11) and Jacobson (12) estimated air quality and associated mortality impacts of using various fuels to power vehicles, including various sources of hydrogen for fuel cell vehicles and different sources of electricity for battery electric vehicles. The primary objective of this study is to determine how ozone concentrations in the Denver, CO metropolitan area might have been different had PHEVs replaced conventional gasoline vehicles in July 2006, given the existing configuration of EGUs in the area. Denver represents an interesting case study for exploring potential PHEV impacts because it violates the federal air quality standard for ozone and because several large EGUs are located in or near the urbanized area. A unit commitment and dispatch model was used to model the charging of the PHEVs and the response of the power plant fleet in the Public Service Company of Colorado (PSCo) service territory. The estimates of generation at individual power plants and changes in gasoline usage were used to estimate changes in NOx and VOC emissions for use in ozone modeling. Though not important for ozone, changes in power plant emissions of sulfur dioxide and primary particulate matter were also estimated. Tailpipe emission reductions associated with PHEV use are assumed to vary in proportion to the reduction in gasoline consumption. Emissions from EGUs were fit using a piecewise linear model of the NOx emissions rate as a function of the heat input, which is output from the unit commitment and dispatch model. Vehicle and power plant emissions were processed for input to the chemistry and transport model and merged with emissions for other sources using the Sparse Matrix Operator Kernel Emissions (SMOKE) modeling system (13). The CAMx model was used to simulate the resulting effects on ozone concentrations. The sensitivity to different PHEV penetration levels and charging scenarios was also considered. Separate responses to across-the-board reductions in VOC and NOx emissions were also estimated to help interpret the results of the PHEV air quality analysis. As noted, this study assumed PHEVs replace light-duty vehicles and affect electricity generation from the power plant fleet as they existed in 2006. Although characteristics of both the vehicle and power plant fleets are likely to change before significant penetration of PHEVs occurs, the hypothetical case provides insight into the potential effects of fuel and vehicle substitution. Current fleet characteristics provide a plausible baseline for this exploratory analysis without introducing additional uncertainties from projecting future fleet characteristics. This study advances understanding of the ozone response to PHEV penetration in several ways. It models a single urban area with a finer grid resolution (4 km) than has been considered in previous PHEV studies, and also incorporates

plume-in-grid treatment for improved tracking of plume dispersion. The Denver metropolitan area has several power plants and no area-wide cap on NOx emissions. The ozone response to PHEV penetration was modeled under a wide range of meteorological conditions, as experienced during July 2006. The NOx emissions from this model reflect significant differences in NOx emission rates between partload and full-load operation of individual EGUs which were not considered in other studies that used a constant emission factor for different categories of units. A range of penetration scenarios was tested to understand the sensitivity of the ozone response to the level of PHEV penetration.

Methods Unit Commitment and Dispatch Model. Unit commitment and dispatch models are optimization tools used to determine the least expensive way to supply the load within the constraints of the electric power system. A unit commitment model and dispatch model developed by Sioshansi and Denholm (8) was modified for the PSCo service territory and used for this study. This model uses a 48-h planning horizon to commit the units to serve the hourly load. The model solves a mixed integer program to minimize the costs of producing electricity and operating (driving, charging, and discharging) the PHEV fleet, including fuel, operation and maintenance, no-load, startup, gasoline, and battery replacement costs. Constraints for conventional generators include minimum and maximum generation and on and off times for each unit, and spinning and nonspinning reserves for the system. Detailed constraints and efficiencies were provided for each unit by Xcel Energy (14). Xcel also provided historical information on electricity demand for 2006. The PSCo system has 7.5 GW of capacity, with coal-fired units representing about 35% of the system capacity, and combined cycle natural gas units representing 45% of the capacity. Combustion turbine natural gas units provide approximately 20% of the capacity; they are primarily used during times of peak demand. The system includes 324 MW of pumped hydroelectric storage capacity. Wind generates less than 2% of the total electricity in the modeled system. The PHEV fleet size is measured as the percentage penetration into the light-duty gasoline vehicle market of PSCo customers in 2006. For example, 30% penetration means that 30% of light-duty gasoline cars and trucks are replaced with PHEVs. In addition to the base case (no penetration of PHEVs), five PHEV scenarios were modeled with the unit commitment and dispatch model. In chargeonly (C-O) scenarios, vehicles can charge whenever they are parked for the entire hour. In vehicle-to-grid (V2G) scenarios, in addition to charging, vehicles can provide power back to the grid for supplying load or meeting reserve requirements. Modeled scenarios include 30% penetration C-O and V2G cases, 100% penetration C-O and V2G cases, and a 100% V2G scenario where charging occurs only between 6 p.m. and 9 a.m. The PHEVs in the model had battery storage capacity of 9.4 kWh. The simulated vehicles were assumed to be operated in the charge-depleting mode (using electricity as the primary fuel) until the state of charge reached 30% of capacity, at which point they were operated in chargesustaining mode (using gasoline as the primary fuel). The driving patterns that determine battery depletion and charging availability were obtained from a household driving survey conducted by the East-West Gateway Coordinating Council (15) that yielded 227 vehicle profiles, which were assumed to be equally distributed among 1.72 million vehicles. Electric efficiency during charge-depleting mode ranged between 210 and 415 Wh/mi (median ) 269 Wh/mi), whereas gasoline efficiency in charge-sustaining mode ranged from 17 to 42 miles per gallon (median ) 37 mpg) (16). Average daily gasoline prices for all grades of gasoline in Colorado during VOL. 44, NO. 16, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6257

July 2006 were obtained from the Energy Information Administration, and ranged from $2.82 to $3.01 per gallon. The Supporting Information includes details on the objective function and constraints for the mixed integer program and plots of the hourly load curve and fuel types used for electricity generation during four modeled days (Figures S1-S6). Emission Changes. Base case pollutant emissions were processed for use in the air quality model with the SMOKE model. The biogenic, area, and mobile source inventories were provided by the National Park Service (17) based on an updated version of the Western Regional Air Partnership (WRAP) inventories from 2002 used for the Rocky Mountain Atmospheric Nitrogen and Sulfur (RoMANS) study. The point source emissions inventories are based on Continuous Emission Monitor (CEM) data from EPA and inventories used for the Colorado State Implementation Plan for ozone, provided by ENVIRON (18). The EPA’s MOBILE6 model was used to estimate base case mobile source emission factors. Gasoline vehicle emissions of NOx and VOCs depend on operating conditions including engine temperature and load. PHEV emissions will depend on how and when the internal combustion engine is used. For this study (and previous studies, e.g., refs 8-10), emission reductions associated with PHEV use are assumed to vary in proportion to the reduction of gasoline consumption. This could underestimate PHEV emissions if the gasoline engine is routinely cold started, or it could overestimate PHEV emissions if the engine never starts for most trips and cold starts are avoided. Because PHEVs are not yet in commercial production, it is difficult to assess this assumption. PHEV penetration is assumed to be limited to PSCo customers, and mobile source emissions are reduced for each county in Colorado based on the percentage of that county’s residents who are PSCo customers. Upstream emissions from sources that produce, transport, or store gasoline (as identified by ref 9) are reduced by the same percentage for each county. Changes to emissions from EGUs are based on the change in generation at each unit between the PHEV scenarios and the base case. Because power plants emit such small quantities of VOCs, NOx is the only power plant pollutant relevant to this study for which emissions are changed for the PHEV scenarios. The EPA’s National Emissions Inventory estimates that 0.4% of VOC emissions in Colorado are from electricity generation. Ninety-four percent of the NOx emissions from EGUs in the PSCo system are from power plants with CEMs. For these units, hourly CEM data from July 2006 were used to create a piecewise linear model to fit the NOx emissions rate as a function of the heat input. If there was an obvious change to the slope of the relationship during part-load operation, a split was made by estimating where the slope changed and two separate lines were fit to the data. Examples are included in the Supporting Information (Figures S10-S11). For units that do not have CEMs, a fixed emission factor (lbs NOx per mmBtu heat input) was used. Air Quality Modeling. CAMx version 4.4 was used to model air pollutant concentrations. CAMx solves the advectiondiffusion equation for a three-dimensional gridded domain. Meteorological fields produced using the Mesoscale Meteorological Model (MM5), and other input files for June and July 2006 were provided by the National Park Service RoMANS study (17). The nested grid structure includes a 36-km grid over the U.S., southern Canada, and northern Mexico, with a 12-km grid over the Rocky Mountain region and a 4-km grid over Colorado. Nineteen vertical layers in the model have interface heights between 18 m and 13 km above the ground. Boundary conditions were calculated from the Model for Ozone and Related Chemical Tracers (MOZART) version 4 (19). A 15-day model spin-up period was run prior to the July 2006 episode. 6258

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 16, 2010

CAMx was run using the Carbon Bond IV (CB-IV) chemistry mechanism to represent gas-phase chemical reactions. The Bott advection scheme was used and the CMC solver was used to integrate the CB-IV chemistry equations. The IRON Plume-in-Grid (PiG) model was used to track concentrated plumes from point sources. The Decoupled Direct Method (DDM) (20) was used to estimate the sensitivity of modeled ozone concentrations to independent changes in NOx and VOC emissions to diagnose whether conditions were limited by NOx or VOC emissions and assist in interpreting the PHEV responses. Emission sensitivities were estimated separately for changes within Colorado and the rest of the modeling domain (“outside Colorado”). Because the unit commitment and dispatch model results were so similar for the charge-only and V2G scenarios, only the V2G scenarios were modeled using the air quality model. In addition to the scenarios previously discussed, two scenarios were run to understand the sensitivity of the results to base case VOC emissions. Because the model underestimated VOC concentrations compared to measurements (Figures S18-S19), the base case and 100% penetration V2G scenario were run assuming doubled VOC emissions from on-road mobile sources in Colorado. The differences between these two scenarios are compared to the differences between the original 100% penetration V2G scenario and the base case to estimate the sensitivity of the results to on-road mobile source VOC emissions.

Results and Discussion Electricity Generation and Gasoline Consumption. The base case electricity demand in the PSCo system varies between approximately 3000 and 5700 MW during the July episode, depending primarily on the time of day. Figures S1-S6 in the Supporting Information show the fuel types of the units used to supply the electricity for the different scenarios for four representative days. Coal provides base load generation; there are typically only a few hours per day with any spare capacity available from the coal units. Natural gas combined cycle and combustion turbine units are projected to provide most (89-97%, depending on the scenario) of the additional electricity required to charge PHEVs. Table 1 displays some of the results of the unit commitment and dispatch modeling, including gasoline and electricity consumption. To supply PHEVs, electricity production is estimated to increase by about 4% for the 30% penetration scenarios and 14% for the 100% penetration scenarios. Gasoline consumption declines by about 19% in the 30% penetration scenarios and 53-64% in the 100% penetration scenarios. The incremental NOx emission factor is lower than the average emission factor for the base case because the incremental electricity is supplied primarily by relatively clean natural gas units. Emission Changes. On-road mobile sources and EGUs respectively emit approximately 25% and 21% of the total NOx emissions in the Denver metro area. In addition to the light-duty gasoline vehicles that were replaced by PHEVs in the scenarios, on-road mobile sources include diesel and heavy-duty gasoline vehicles, which were assumed not to be affected. Table 2 shows the results of the emissions modeling for NOx and VOCs in Colorado. NOx emissions from EGUs are projected to be 1.9-3.3 tons per day (tpd) (1.2-2.2%) higher than in the base case, depending on the scenario. The model projects NOx emissions from on-road mobile sources to be reduced by 8.1-27.2 tons per day (5-16%). The total NOx reduction is less than 4% in all scenarios. The NOx emission factor (total NOx emitted by the vehicle or EGU per distance driven) is reduced by 73-88% when an average PHEV is operated using electricity. Although almost all of the on-road mobile source NOx reductions occur within 100 km of Denver, most of the NOx increases from EGUs occur more than 100 km away from the city. The percentage reduction

TABLE 1. Unit Commitment and Dispatch Model Results for EGUs in the PSCo System and Light Duty Gasoline Vehicles in the PSCo Service Area (Daily Average for July)

Case

electricity consumptiona (MWh/day)

NOx emission factora (lbs/MWh)

percent of gasoline VMTsb replaced by electricity

gasoline consumption (gallons/day)

base case (average) 30% C-O (incremental) 30% V2G (incremental) 100% C-O (incremental) 100% V2G (incremental) 100% V2G, night charging (incremental)

99,500 4200 4000 13,900 13,800 11,500

2.95 1.05 0.95 0.47 0.46 0.43

0% 19.3% 18.9% 64.1% 63.4% 53.3%

1,810,000 -349,000 -343,000 -1,160,000 -1,150,000 -964,000

a Base case values are system averages, PHEV scenario values are incremental values for the additional electricity produced or gasoline consumed compared to the base case. b VMTs ) vehicle-miles traveled.

TABLE 2. Average Base Case NOx and VOC Emissions from EGUs and On-Road Mobile Sources in Colorado during July, 2006 and Changes in Emissions for PHEV Scenarios

base case (total) 30% C-O (incremental) 30% V2G (incremental) 100% C-O (incremental) 100% V2G (incremental) 100% V2G, night charging (incremental)

NOx from electric generating units (tpd)

NOx from on-road mobile sources (tpd)

148 +2.2 (1.4%) +1.9 (1.2%) +3.3 (2.2%) +3.2 (2.1%) +2.5 (1.7%)

176 -8.3 (-4.7%) -8.1 (-4.6%) -27.4 (-16%) -27.2 (-15%) -22.8 (-13%)

total NOx (tpd)

NOx emission factor reductiona

VOCs from on-road mobile sources (tpd)

709 -7.3 (-1.0%) -7.2 (-1.0%) -25.0 (-3.5%) -24.8 (-3.5%) -20.9 (-3.0%)

N/A 73% 77% 88% 88% 88%

236 -17.3 (-7.3%) -17.0 (-7.2%) -57.6 (-24%) -57.0 (-24%) -47.9 (-20%)

a NOx emission factor reduction is the difference in total NOx emitted per distance driven between using gasoline or electric power.

FIGURE 1. Base case NOx emissions (left) and NOx emission changes (right), 100% penetration V2G scenario minus base case, 15:00 MST, July 14. of VOC emissions from on-road mobile sources is larger than the percentage reduction of NOx emissions from on-road mobile sources because light-duty gasoline vehicles emit a larger percentage of the on-road mobile VOC emissions compared to NOx emissions. However, the VOC emission reduction is less than 2% of total VOC emissions for all cases because of other sources of VOCs, especially biogenic sources. Figure 1 shows the base case NOx emissions and changes in total NOx emissions for the 100% penetration V2G case for a subset of the 4-km modeling domain. The maps show an area approximately 190 km by 220 km centered on Denver and extending from Fort Collins in the north to Colorado Springs in the south (Figure S7). The area around Denver shows the largest reductions in NOx emissions. This is where the population and roads are the densest and NOx emissions

are highest in the base case. As expected based on driving patterns, emissions reductions from vehicles in the PHEV cases occur primarily from 06:00 to 18:00 MST. In contrast, emissions increases from power plants due to vehicle charging occur mainly from midnight to 06:00 MST. The dispatch modeling results indicate that changes in SO2, PM10, and PM2.5 emissions from EGUs lead to increases of less than 1% of the total emissions inventories. The increase in SO2 emissions is modest because generation from coalfired power plants did not change significantly in the PHEV scenarios. PM emission changes were modest because EGUs contribute little to primary PM emissions in Colorado (less than 10% according to EPA’s National Emissions Inventory). Air Quality Changes. The base case model output represents air quality in the Denver metro area reasonably VOL. 44, NO. 16, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6259

FIGURE 2. Base case (left) and difference in (right) maximum 8-h average ozone concentration for the 100% penetration V2G scenario for July 14. well. The base case results were compared to hourly ozone concentrations observed at 17 locations along the Colorado Front Range for the period from July 1 to July 31, 2006. Comparisons for four representative monitoring sites are provided in the Supporting Information (Figures S10-S13) with additional results presented by Brinkman (21). The model represents the spatial and temporal patterns of ozone concentrations well at most time periods in most areas. However, ozone concentrations at the monitoring locations west of Denver were typically underpredicted (e.g., Figure S12). On most days, the modeled area of elevated ozone was too small and should have been centered farther to the west of the metropolitan area. On many days during the episode, the model underpredicted nighttime ozone concentrations at some of the urban and suburban monitoring locations near NOx sources (e.g., Figure S10). Consequently, the nighttime changes in ozone concentrations due to PHEV penetration should be viewed with caution. The nighttime underprediction could be caused by insufficient vertical mixing in the model, an overestimation of nighttime NOx emissions, or other possible causes. The mean normalized bias for all 17 monitoring locations was -13% during the day (6 a.m.-6 p.m.) and -23% at night (only pairs with observed concentrations greater than 40 ppb were included in the calculation). Modeled VOC/NOx ratios near Denver were underestimated by an average of 27% at the two locations where these ozone precursors were monitored in July 2006 (Figures S14-S15). The Decoupled Direct Method was used to estimate the sensitivity of modeled ozone to uniform percentage changes in VOC and NOx emissions from three different areas: the Denver metro area, the rest of Colorado, and the rest of the model domain covered by the 36-km grid. For comparison, results were scaled to emissions changes of 10% across each area. Full results are available from ref 21; representative figures are available in the Supporting Information (Figures S16-S20). During July, most of central Colorado displayed relatively strong sensitivity to NOx emissions. Areas up to 50 km from central Denver often displayed relatively strong sensitivity to local VOC emissions, although the size of this area varied by day. The most intense areas of VOC sensitivity were often accompanied by an area projected to have reduced ozone concentrations in response to local NOx emission increases. Although VOC emissions from outside Colorado typically had little effect on the modeled ozone concentrations in the focus area, NOx emissions outside Colorado were more influential. On some days, a 10% increase in NOx emissions outside Colorado caused a larger increase in modeled ozone concentrations in the focus area than a 10% increase in NOx 6260

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 16, 2010

emissions within the state. Increased local NOx emissions typically reduce nighttime ozone concentrations in the focus area. The largest positive sensitivity to increased NOx emissions is seen at the time of peak ozone. The results of the PHEV scenario air quality analysis suggest that if PHEVs replaced current vehicles and were charged with electricity from PSCo’s existing power plant fleet, PHEV penetration would reduce daily peak ozone concentrations in the Denver metro area. Ozone reductions in the PHEV scenarios tend to occur at times and places with moderate or high ozone and in areas outside of the VOCsensitive areas close to downtown Denver. The largest ozone reductions occur at times and places with the highest ozone concentrations; the largest increases occur at times and places with relatively low ozone concentrations (usually less than 35 ppb). Modeled ozone increases due to PHEV penetration typically occur in VOC-sensitive areas. Although VOC emissions in these areas are reduced (which by itself would probably reduce ozone concentrations), total VOC emissions in the focus area are reduced by a smaller percentage compared to NOx emissions. The net effect creates ozone increases in some of the VOC-sensitive areas of the model. Figure 2 shows an example of the maximum 8-h average ozone concentration difference between the 100% penetration V2G scenario and the base case for July 14, a day with a typical response to PHEVs. On this date, modeled maximum 8-h average concentrations of up to 84 ppb occurred north of Denver and coincided with the area of maximum reductions. Small increases in ozone occurred in central Denver where the maximum 8-h average ozone concentrations were approximately 65 ppb. As illustrated in the Supporting Information (Figures S21-S26), some days with relatively low peak ozone concentrations showed smaller reductions and larger increases in maximum 8-h average ozone in the PHEV scenarios, while other days with relatively high peak ozone concentrations showed larger reductions over a larger area and almost no increases. The response to different PHEV penetration and charging scenarios was nearly linear with respect to the fraction of VMTs powered by electricity (21). For example, the ozone response in the 30% penetration V2G scenario is approximately 30% of the response in the 100% penetration scenario. Figures S27-S32 in the Supporting Information illustrate how the PHEV scenarios impact ozone concentrations diurnally, showing base case concentrations and the response to 100% PHEV penetration (V2G) at 03:00, 09:00, and 15:00 MST on July 14. The sensitivity of the ozone response to base case onroad mobile source VOC emissions was tested by doubling the VOC emissions from on-road mobile sources in the base

TABLE 3. Change in the Number of Grid Cell-Hours above Ozone Concentration Thresholds

a

Ozone concentration threshold

60 ppb

75 ppb

85 ppb

95 ppb

base case (grid cell-hours) 30% V2Ga 100% V2Ga 100% V2G, night charginga double VOC base case (grid cell-hours) double VOC, 100% V2Gb

171000 -1.1% -4.1% -3.5% 174000 -4.8%

12100 -5.4% -20.5% -17.1% 13900 -23.9%

2130 -9.4% -29.5% -24.4% 2900 -36.4%

353 -14.2% -45.3% -40.8% 696 -59.9%

Change compared to the base case.

b

Change compared to the double VOC base case.

case and the 100% penetration V2G scenario. With doubled on-road mobile source VOC emissions, the maximum 8-h average ozone concentrations would increase by less than 3 ppb over a small VOC-sensitive area. The modified response to PHEV penetration depends on the local conditions in the model. With doubled on-road mobile VOC emissions, the ozone increases in the PHEV scenario are reduced in magnitude and extent, and ozone reductions are increased. For plots of the differences, see Figures S33-S38. To summarize the results, Table 3 shows the change in the number of grid cell-hours above ozone concentration thresholds for the different scenarios.

Discussion The unit commitment and dispatch model estimates that the electricity used to charge PHEVs in Denver would primarily come from natural gas power plants. The response of the electric power system to V2G charging and discharging is very similar to the charge-only scenarios for any given penetration because the system already has enough lowcost spinning reserves (partly from pumped hydroelectric storage) and there is not enough difference in marginal production cost between different times of day to justify discharging PHEV batteries to the grid. The models estimate that emission reductions from on-road mobile sources occur primarily in heavily populated areas where vehicles are driven more, and NOx increases from EGUs occur primarily (although not exclusively) in less populated areas. There are several major power plants in populated areas; however, most of these plants do not change output significantly due to PHEV penetration in the model. While the ozone response is mixed, PHEV penetration reduced ozone at times and locations with the highest base case ozone concentrations. The peak maximum 8-h average ozone concentration is typically reduced by 2-3 ppb (in the case of 100% PHEV penetration) and slightly less on days with lower ozone concentrations. Small areas of ozone concentration increases are projected in central Denver. The response of ozone to PHEV penetration scales approximately proportionally to the fraction of gasoline VMTs replaced by electricity. In summary, the penetration of PHEVs in the scenarios modeled led to modest ozone benefits in the Denver metropolitan area. While the detailed case study for Denver complements national-scale assessments such as that conducted by EPRI (9), results from this study may not extend to other areas due to Denver’s distinguishing characteristics. The marginal generating units in the PSCo system have significantly lower NOx emission rates compared to units that would provide marginal electricity to charge PHEVs in many other areas. Modeled ozone concentrations in the Denver area are usually sensitive to NOx emissions, although there is typically a small area in the urban core that is sensitive to VOC emissions. The Denver metropolitan area is relatively isolated from other cities, but the air quality model estimates that proportional changes to NOx emissions from outside Colorado produce similar ozone differences compared to proportional changes

in local NOx emissions. EPRI (9) presented results for the fourth highest annual 8-h average ozone concentration; the changes they found near Denver are similar to those obtained in the present study on high-ozone days in the 30% penetration scenario (up to 1 ppb), which assumed a similar fraction of VMTs powered by electricity. This study for Denver focused on impacts on ozone concentrations, due to Denver’s nonattainment status and interest in the potentially nonlinear responses this pollutant can exhibit. For other areas, where PM2.5 concentrations are higher than in Denver and where coal-fired power plants may contribute more of the marginal electricity supply, modeling effects of PHEV use on PM2.5 may also be warranted, especially if emissions caps are not in place to mitigate this impact. In future work, additional research is needed to refine estimates of PHEV emissions, including understanding the importance of cold-start conditions. Further research is also needed to forecast the timing of PHEV penetration and better anticipate the characteristics of the power plant and vehicle fleets that might be in place when significant PHEV use might occur. Such projections will have high uncertainty, because without having PHEVs in mass production it is difficult to estimate their emission characteristics and added upfront costs, including the cost of battery production. Over the longterm, it is possible that significant PHEV penetration could affect how the generation fleet develops, possibly by leading to more base load coal capacity. Finally, future studies should pursue dispatch modeling that extends beyond the service territory of a single electric power utility to examine the effects of more widespread introduction of PHEVs.

Acknowledgments This research was supported by the University of Colorado Renewable and Sustainable Energy Institute. We thank Keith Parks and Xcel Energy, Mike Barna, Ralph Morris, Ramteen Sioshansi, and Easan Drury for their contributions.

Supporting Information Available The objective function and additional detail on the unit commitment and dispatch model, generation curves by electric generating unit type, emission curves of selected electric generating units, air quality model base case performance evaluation, emission time series plots by scenario, and ozone concentration difference plots for selected dates and times. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Annual Energy Outlook 2009 with Projections to 2030; DOE/ EIA-0383; U.S. Department of Energy, Energy Information Administration: Washington, DC, March 2009. (2) Stephan, C. H.; Sullivan, J. Environmental and energy implications of plug-in hybrid-electric vehicles. Environ. Sci. Technol. 2008, 42 (4), 1185–1190. (3) Bell, M. L.; McDermott, A.; Zeger, S. L.; Samet, J. M.; Dominici, F. Ozone and short-term mortality in 95 US urban communities, 1987-2000. JAMA, J. Am. Med. Assoc. 2004, 292 (19), 2372–2378. VOL. 44, NO. 16, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6261

(4) Parks, K.; Denholm, P.; Markel, T.; U.S. Environmental Protection Agency. National Ambient Air Quality Standards for Ozone, Final Rule. Fed. Regist. 2008, 73 (60), 16436–16514. (5) Full Fuel Cycle Assessment: Well-to-Wheels Energy Inputs, Emissions, and Water Impacts; California Energy Commission, June 2007. (6) Gaines, L., Burnham, A., Rousseau, A., Santini, D. Sorting Through the Many Total-Energy-Cycle Pathways Possible with Early Plug-In Hybrids; Electric Vehicle Symposium 23, Anaheim, CA, December 2-5, 2008. (7) Costs and Emissions Associated with Plug-In Hybrid Electric Vehicle Charging in the Xcel Energy Colorado Service Territory; NREL/TP-640-41410; National Renewable Energy Laboratory: Golden, CO, 2007. (8) Sioshansi, R.; Denholm, P. Emissions Impacts and Benefits of Plug-In Hybrid Electric Vehicles and Vehicle-to-Grid Services. Environ. Sci. Technol. 2009, 43 (4), 1199–1204. (9) Environmental Assessment of Plug-In Hybrid Electric Vehicles, Vol. 2: United States Air Quality Analysis Based on AEO-2006 Assumptions for 2030; Electric Power Research Institute: Palo Alto, CA, 2007. (10) Thompson, T.; Webber, M.; Allen, D. T. Air quality impacts of using overnight electricity generation to charge plug-in hybrid electric vehicles for daytime use. Environ. Res. Lett. 2009, 4, 014002. (11) Colella, W. G.; Jacobson, M. Z.; Golden, D. M. Switching to a US hydrogen fuel cell vehicle fleet: The resultant change in emissions, energy use, and greenhouse gases. J. Power Sources 2005, 150, 150–181. (12) Jacobson, M. Z. Review of solutions to global warming, air pollution, and energy security. Energy Environ. Sci. 2009, 2 (2), 148–173. (13) SMOKE v2.3 User’s Manual; Carolina Environment Program: Chapel Hill, NC, 2006.

6262

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 16, 2010

(14) Parks, K. Xcel Energy, Denver, CO, Personal Communication, October 2008. (15) Household Travel Survey: Final Report of Survey Results; EastWest Gateway Coordinating Council: St Louis, MO, 2003. (16) Gonder, J.; Markel, T.; Thornton, M.; Simpson, A. Using global positioning system travel data to assess real-world energy use of plug-in hybrid electric vehicles. Transp. Res. Rec. 2007, (2017), 26–32. (17) Malm, W., Collett, J., Barna, M., Beem, K., Carrico, C., Gebhart, K., Hand, J., Levin, E., Rodriguez, M., Schichtel, B., Schwandner, F. RoMANS: Rocky Mountain Atmospheric Nitrogen and Sulfur Study Report; National Park Service: Washington, DC; October, 2009. (18) Morris, R. ENVIRON, Novato, CA, Personal Communication, April 2008. (19) Pfister, G. G.; Emmons, L. K.; Hess, P. G.; Lamarque, J. F.; Thompson, A. M.; Yorks, J. E. Analysis of the Summer 2004 ozone budget over the United States using Intercontinental Transport Experiment Ozonesonde Network Study (IONS) observations and Model of Ozone and Related Tracers (MOZART-4) simulations. J. Geophy. Res.-Atmos. 2008, 113, Article Number: D23306. (20) Dunker, A. M.; Yarwood, G.; Ortmann, J. P.; Wilson, G. M. The decoupled direct method for sensitivity analysis in a threedimensional air quality model - Implementation, accuracy, and efficiency. Environ. Sci. Technol. 2002, 36 (13), 2965–2976. (21) Brinkman, G. Air Pollutant Sources and Exposure: Organic Tracers for Source Identification of Exposure to Fine Particles and Effects of Plug-in Hybrid Electric Vehicles on Ozone Concentrations in Colorado. Ph.D. Dissertation, University of Colorado, Boulder, CO, 2009.

ES101076C