- Email: [email protected]
Emissions implications of downscaled electricity generation scenarios for the western United States
Energy Policy 109 (2017) 601–608
Contents lists available at ScienceDirect
Energy Policy journal homepage: www.elsevier.com/locate/enpol
Emissions implications of downscaled electricity generation scenarios for the western United States Rene Nsanzinezaa, Matthew O’Connellb, Gregory Brinkmanb, Jana B. Milforda, a b
MARK
⁎
Department of Mechanical Engineering, University of Colorado, 1111 Engineering Drive, Boulder, CO 80309-0427, USA Strategic Energy Analysis Center, National Renewable Energy Laboratory, 15013 Denver West Parkway, Golden, CO 80401, USA
A R T I C L E I N F O
A B S T R A C T
Keywords: Electricity dispatch Air quality Emissions fees Nitrogen oxides Natural gas Renewable energy
This study explores how emissions from electricity generation in the Western Interconnection region of the U.S. might respond in circa 2030 to contrasting scenarios for fuel prices and greenhouse gas (GHG) emissions fees. We examine spatial and temporal variations in generation mix across the region and year using the PLEXOS unit commitment and dispatch model with a production cost model database adapted from the Western Electricity Coordinating Council. Emissions estimates are computed by combining the dispatch model results with unitspecific, emissions-load relationships. Wind energy displaces natural gas and coal in scenarios with relatively expensive natural gas or with GHG fees. Correspondingly, annual emissions of NOx, SO2, and CO2 are reduced by 20–40% in these cases. NOx emissions, which are a concern as a precursor of ground-level ozone, are relatively high and consistent across scenarios during summer, when peak electricity loads occur and wind resources in the region are comparatively weak. Accounting for the difference in start-up versus stabilized NOx emissions rates for natural gas plants had little impact on region-wide emissions estimates due to the dominant contribution from coal-fired plants, but would be more important in the vicinity of the natural gas units.
1. Introduction Led by Wyoming, the eight states in the Rocky Mountain region (AZ, CO, ID, MT, NM, NV, UT, WY) account for more than 50% of U.S. coal production and 18% of natural gas production (EIA, 2016a). Coal production in the region peaked at over 600 million short tons in 2008 and fell to 480 million tons in 2015 (EIA, 2016a, 2016b). In 2015, the states in the Rocky Mountain region produced 5.1 trillion cubic feet of natural gas, up 6% from the level in 2005, with sharp growth in Colorado offsetting a decline in New Mexico and relatively stable production in Wyoming (EIA, 2016a, 2016c). Since 2010, utilities in the region have completed or announced plans to retire or repower more than 20 coal-fired power plant units. Electricity generation in these states shifted from 63% coal, 19% natural gas and less than 1% wind in 2005, to 52% coal, 22% natural gas and 5% wind in 2014 (EIA, 2016d). Shifts to natural gas and renewable energy are expected to continue in the western U.S., with significant implications for the region's air quality. Electricity generation and energy production activities are
major sources of nitrogen oxides (NOx), sulfur oxides, volatile organic compounds (VOC) and greenhouse gases (GHG). Except for Idaho and Montana, all of the states in the Rocky Mountain region include areas that have been designated nonattainment for the 2008 National Ambient Air Quality Standard for ozone or have recent ozone values above the 2015 revised standard (EPA, 2016a), so emissions of ozone precursors – NOx and VOCs – are of particular concern. Understanding how the changing energy landscape might affect future emissions and air quality in the Rocky Mountain region requires a combination of energy and air quality models that bridge across hourly to decadal time scales. The objective of this study is to examine the sub-regional spatial patterns and sub-annual temporal patterns of electricity sector emissions within the Rocky Mountain region that might result from contrasting scenarios for future natural gas prices and greenhouse gas mitigation policies. The study builds on prior work by McLeod et al. (2014), who used the U.S. Environmental Protection Agency's (EPA) nine-region MARKAL energy system model to examine how annual
Abbreviations: CSP, Concentrating Solar Power; EIA, Energy Information Administration; EPA, Environmental Protection Agency; ERCOT, Electric Reliability Council of Texas; GHG, greenhouse gases; NREL, National Renewable Energy Laboratory; PAWY, PacifiCorp East – Wyoming; PSCO, Public Service Company of Colorado; PV, photovoltaics; RPV, rooftop photovoltaics; TEPPC, Transmission Expansion Planning Policy Committee; VOC, volatile organic compounds; WACM, Western Area Power Administration – Colorado-Missouri; WECC, Western Electricity Coordinating Council ⁎ Corresponding author. E-mail addresses: [email protected] (R. Nsanzineza), [email protected] (M. O’Connell), [email protected] (G. Brinkman), [email protected] (J.B. Milford). http://dx.doi.org/10.1016/j.enpol.2017.07.051 Received 20 February 2017; Received in revised form 6 June 2017; Accepted 23 July 2017 0301-4215/ © 2017 Published by Elsevier Ltd.
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
meteorological conditions are conducive to ozone formation. Fig. A1 in Appendix A illustrates such conditions, showing the correlation between temperature, power plant load and ozone for Denver, CO in July 2006. Recent studies have started to conduct electricity system analyses that provide higher resolution, by using electricity dispatch models or simplified representations of dispatch order to estimate emissions changes in response to altered demand, fuel prices, or policies to encourage renewables (e.g., Hobbs et al., 2010; Brinkman et al., 2010; Thompson et al., 2011; Plachinski et al., 2014). For example, Gilbraith and Powers (2013) applied a dispatch model to simulate residential demand response in New York City, finding that a moderate program could reduce generation from small, relatively inefficient combustion turbines used to meet peak demand, thus reducing NOx and PM2.5 emissions on poor air quality days. Buonocore et al. (2015) applied a dispatch model for the Eastern Interconnection to compare the emissions and corresponding health benefits of incorporating wind, solar, and demand response at six locations in the eastern U.S., illustrating how the benefits depend on the type and location of fossil fuel generation being displaced. Kerl et al. (2015) used a reduced-form air quality model relating emissions to pollutant concentrations and monetized estimates of resulting health effects to generate environmental damage costs for inclusion in a least-cost electricity dispatch algorithm for power plants in Georgia. Including environmental costs in the dispatch algorithm led to shifting generation on some winter days from coal-fired power plants in northern Georgia to a natural gas combined cycle plant near the coast. Pacsi et al. (2013) used the PowerWorld Simulator to estimate how hourly generation from units in the existing Electric Reliability Council of Texas (ERCOT) system depends on natural gas prices, as an input to an assessment of net air quality impacts from natural gas production in the Barnett shale. They estimated NOx emissions for each unit by multiplying hourly generation with the unit's annual average emissions factor. Net impacts on ozone air quality in Texas were modeled using a regional atmospheric chemistry and transport model, CAMx (Comprehensive Air Quality Model with Extensions), adjusting the power plant NOx emissions and estimates of NOx and VOC emissions from local natural gas production in each gas price case. Pacsi et al. (2015) conducted a similar net air quality analysis assuming the natural gas came from the Eagle Ford shale, instead of the Barnett. Compared to prior emissions studies using dispatch modeling, this study is unique in focusing on the Rocky Mountain region, which has abundant wind and solar resources as well as coal and natural gas. We examine relatively broad and self-consistent scenarios for key factors that could shape future electricity generation in the region, covering a range of natural gas prices and greenhouse gas mitigation policies. Unlike other studies that have considered limited adjustments to the current electricity system, we examine dispatch results for a future electricity generating fleet that reflects utilities’ current plans for retiring or repowering coal plants. Lastly, unlike most prior studies, we use detailed unit-specific emissions models that account for load-dependence of emissions rates.
average emissions would respond to these scenarios out to the year 2050, for the entire U.S. and for the Rocky Mountain region as a whole. To begin to examine spatial and temporal patterns within the Rocky Mountain region, the current study “downscales” the electricity sector emissions estimates for this region for a circa 2030 time period. To do this, we simulate the electricity generation mix in selected scenarios at hourly temporal resolution using the PLEXOS unit commitment and dispatch model for the power plant fleet expected to be in place in the Western Interconnection at that time. Hourly variations in hydro, wind and solar resource availability and electricity demand are estimated from historical data, which are then scaled by increased overall demand and renewable electricity capacity for the future scenarios. Emissions are computed by combining unit-specific loads from the dispatch model with load-dependent emissions factors. To our knowledge, this is the first such study for the Rocky Mountain region, and the first study in any U.S. region to provide this level of detail for a future timeframe when significantly expanded capacity for renewable energy could be in place to compete with natural gas and coal. Future work will combine the electricity sector emissions results described here with estimates of emissions changes from upstream energy production, and use these as inputs to a regional-scale chemistry and transport model to study net air quality impacts of natural gas production and use. 2. Background The U.S. EPA, U.S. Energy Information Administration (EIA) and other Department of Energy laboratories have long used least-cost or partial equilibrium energy system or power sector planning models to examine emissions impacts of future electricity generation or broader energy scenarios and to analyze proposed emissions control strategies and other regulations. These models include EIA's National Energy Modeling system (EIA, 2013) used to produce their Annual Energy Outlook; EPA's MARKAL model and nine-region U.S. database (Lenox et al., 2013); the Integrated Planning Model (EPA, 2014) used in regulatory analyses such as those for EPA's Cross-State Air Pollution Rule (EPA, 2016); and the National Renewable Energy Laboratory's Regional Energy System Deployment (ReEDS) model (Eurek et al., 2016; Cole et al., 2016). Researchers have used these and similar models to examine a range of scenarios and policy proposals. For example, Brown et al. (2013, 2017) applied MARKAL with EPA's nine-region database to examine how adding damage-based fees in the electric power and other U.S. energy sectors would alter emissions. Trail et al. (2014) used emissions from EPA's MARKAL reference case to examine how air quality across the U.S. might change by 2050. Thompson et al. (2014, 2016) used the U.S. Regional Energy Policy model to estimate conventional air pollutant emissions responses to greenhouse gas mitigation strategies. In work that forms the basis for the current study, McLeod et al. (2014) modified EPA's MARKAL model with updated information on costs of renewable energy and emissions from oil and gas production, then used the revised model to examine how contrasting scenarios for natural gas supply and demand, constraints on the electricity generation mix, and GHG fees might affect energy system emissions for the U.S. and the Rocky Mountain region out to the year 2050. These studies all focus on projecting annual average emissions changes at the state, regional, or national scale. In studies that estimate air quality impacts from the energy and emissions modeling results, the projected annual changes in emissions are typically used to scale existing emissions inventories up or down, assuming the location and subannual timing of emissions will not change (e.g., Thompson et al., 2014, 2016). Higher resolution in emissions projections is critical for examining how alternative economic or policy scenarios might influence emissions during specific time periods that are prone to air quality problems (Hobbs et al., 2010; Krieger et al., 2016), e.g., on hot summer days when electricity demand is high, wind availability tends to be low, and
3. Methods 3.1. Scenarios The natural gas price and emissions fee scenarios and corresponding renewable energy build outs considered in this study were developed by McLeod et al. (2014), using the MARKAL model with the EPA U.S. nineregion database. The MARKAL model finds the least cost means to satisfy future end use demand in the industrial, commercial, residential, and transportation sectors, under specified constraints including limits on fuel supplies and on rates of capacity expansion and introduction of new technology. The EPA database available at the time was developed to match projections from Energy Information Administration's 2012 Annual Energy Outlook. McLeod et al. (2014) modified the 2012 602
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
Anido, 2016; Novacheck and Johnson, 2017). What this means is that PLEXOS will simulate the unit commitment (decisions to start up or shut down) and dispatch (adjusting generation levels) of electrical generation units to match grid energy demand for the lowest cost possible solution, subject to a suite of constraints. These constraints include transmission line flow limits, minimum and maximum generation levels, generation ramping limits, and maximum or minimum energy output per month for hydro plants. Costs include fuel cost, start and shutdown costs, variable operation and maintenance costs, and emissions fees (if applicable). The PLEXOS software was run at the National Renewable Energy Laboratory (NREL) using a modified version of the Western Electricity Coordinating Council's (WECC) Transmission Expansion Planning Policy Committee's (TEPPC) 2024 Common Case production cost model database (WECC, 2015). This database represents the expected 2024 Western Interconnection electrical grid and generation fleet. We made minor changes to the 2024 TEPPC database in order to implement the desired scenarios. The database contains 4836 generators and 43 individual load regions, with a total of 8349 transformers, 16,847 transmission lines, and 19,782 individual nodes. The generators have unique properties specified such as fuel type, heat rate, maximum capacity, minimum generation levels, operating cost, ramp rate limitations, and minimum run time and down time. Hourly load profiles are based on the actual hourly loads from 2006, scaled to expected 2024 load magnitude. Hourly load profiles are provided in the TEPPC 2024 Common Case database and estimate demand in 2024 for each balancing area, with a total interconnection peak demand of 178.9 GW occurring on July 24th at 5:00 P.M. PST. For purposes of this study, we assume the 2024 fleet configuration and demand estimates are representative for a circa 2030 timeframe. We used the TEPPC database as the best information available, and use the phrase “circa 2030” as the midpoint between the TEPPC load and fossil fleet infrastructure projections and the 2035 MARKAL results for renewables build out. The generation capacity of the database, broken down into individual areas and excluding variable generation renewables, is shown in Appendix B, Table B1. Variable generation technologies such as wind and solar PV are not shown because each case studied has a different renewable portfolio build out. As noted above, the renewable build outs are taken from McLeod et al. (2014). Note that use of the 2024 TEPPC database results in more coal capacity and coal generation than was found in the prior study with MARKAL, for which coal plant retirements in the region were further advanced. Renewable generation resources were developed for the model through a combination of existing projects, near-term projects under development, and estimated future projects. Generation profiles for these resources are based on 2006 meteorological conditions and were developed by NREL for TEPPC. Wind and solar generation in the PLEXOS simulations are based on these profiles, which are provided to the PLEXOS simulations as inputs. Hourly generation profiles are available for the entire year for all the wind and solar sites, and vary for each case. The database contains 417 wind generators and 596 solar generators, which are separated into three categories. There are 31 concentrated solar power (CSP) generators, 45 rooftop photovoltaic (RPV) sites representing many installations, and 520 utility scale photovoltaic (PV) installations. Total available generation from solar is held constant across scenarios at 78.6 TWh. Available generation from wind ranges from 82.7 TWh in the base case to 268 TWh in the GHG fees case. Appendix B Table B2 shows total generation from wind for each scenario, broken down by season. Solving an optimal power flow problem at the individual node level for the entire western interconnection is computationally difficult and would take an excessive amount of time. To decrease computing time, transmission networks within each region are aggregated into a single node, and all generators in the region are also assigned to this same node. Each region's load profile is represented 100% by its respective
version of EPA's nine-region energy system and emissions database to update cost and performance data for electric generating units, including for PV and wind, update renewable portfolio standards to reflect 2013 developments, separate unconventional and conventional natural gas production, and to account for emissions reductions in the oil and gas production sector due to EPA regulations promulgated in 2012. McLeod et al. (2014) modeled seven scenarios, four of which are considered as a starting point for this study. Results were examined at the national scale and specifically for the Rocky Mountain region. The base case reflects fuel supply and cost assumptions for the U.S. and Rocky Mountain region from the 2013 Annual Energy Outlook reference case. Two scenarios examined contrasting assumptions about future natural gas production, again based on AEO 2013, with “cheap gas” assuming relatively abundant supply and low prices, and “costly gas” assuming limited supply and high prices. The other scenario considered here used the base case assumptions for natural gas supplies, but applied GHG fees to CO2 and methane emissions in all sectors, including electricity generation and oil and gas production, based on the Social Cost of Carbon adopted by the Interagency Working Group in May 2013 (IWG, 2013). Fig. 1 is adapted from McLeod et al. (2014) and shows the electricity generation mix in the eight state Rocky Mountain region in the base year, 2010, and as modeled with MARKAL for 2035 in each of the four scenarios. (Results for the U.S. as a whole are given in McLeod et al., 2014.) For the Rocky Mountain region, electricity generation from natural gas and wind increases between 2010 and 2035. Wind displaces natural gas in the costly gas scenario, and displaces natural gas and coal in the scenario with greenhouse gas fees. Electricity generation in the costly gas and GHG fees scenarios is higher than in the base case due to increased electricity exports from the region. The current work uses estimates of renewable generation capacity, natural gas prices, and GHG fees from McLeod et al. (2014), but incorporates them into a unit commitment and dispatch model. Shifting to an hourly dispatch modeling framework provides the needed resolution to understand emissions implications for the electricity generation sector on the relatively short time-scale of air pollution episodes, and also allows consideration of how intraregional transmission constraints and ancillary service requirements might alter the choice between coal, natural gas, and renewables. 3.2. Dispatch modeling We used the PLEXOS unit commitment and economic dispatch model to simulate operation of the western US electric interconnection. PLEXOS is an optimization tool licensed and supported by Energy Exemplar that can solve unit commitment and economic dispatch problems (e.g., Bloom et al., 2016; Brinkman et al., 2016; Martinez-
Fig. 1. Rocky Mountain electricity generation mix, TWh (adapted from McLeod et al., 2014).
603
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
for stabilized (above 40% or 50% load) and low-load conditions. For comparison, we also estimate NOx emissions under the assumption that emissions are a constant multiplier of load across all load levels. Modeling emissions as a function of load and considering start-up emissions explains most of the additional emissions experienced during ramping and cycling operations. Further detail on the treatment of start-up emissions and discussion of their influence is provided in Appendix C. For power plants with pending control requirements that are tighter than those in place in 2011, we calculated an emissions control factor by comparing the maximum allowable rate in pending requirements to the maximum emissions rate in the 2011 dataset, using the averaging period specified in the pending requirements. If the maximum rate in 2011 was less than the pending limit, no adjustment was made to the historical emissions factors. For SO2 emissions, we used 2015 data from EPA's Air Markets Database (EPA, 2016b) to estimate unit specific emissions factors (tons/ MW gross load) for individual coal-fired power plants. These emissions factors were combined with annual electricity generation from the dispatch model results for each unit to estimate annual SO2 emissions. For natural gas power plants, we calculated average emissions factors by state and unit type (combustion turbine or combined cycle). The average emissions rate for natural gas by state was multiplied with the heat input data from the dispatch model to estimate the SO2 emissions from power plants in each of the future scenarios. Using statewide average emissions factors for natural gas power plants has little effect on the results because their SO2 emissions are low compared to those from coal-fired power plants. We estimated CO2 emissions for the future scenarios using emissions rates of 214 lb/MMBtu for coal and 117 lb/ MMBtu for natural gas (EIA, 2016e).
Table 1 Natural gas and coal fuel prices used in PLEXOS ($/MMBtu). Prices shown for the GHG fees case are added to those in the base case. Scenario
Natural gas in CA, OR, and WA
Natural gas elsewhere in Western Interconnection
Coal (price varies by state)
Base Cheap Gas Costly Gas GHG Fees
7.58 4.72 9.81 + 3.86
7.36 4.73 9.66 + 3.86
0.97 to 3.18 0.97 to 3.18 0.97 to 3.18 + 6.03
node. This simplification decreases computational time significantly, although due to the aggregation of lines it also reduces the accuracy of the solutions. In addition to this regional aggregation, all line reactance and resistances are removed and PLEXOS solves a transport flow problem, meaning the solver can decide how much power to send down each line to serve load in each corresponding region as long as line power flow constraints are not violated. The unit commitment and economic dispatch problem, including optimal power flow, was solved for the western interconnection for the four scenarios described previously. Natural gas prices used in PLEXOS are different in each scenario, but coal and liquid fuel (oil, propane, etc.) prices are the same except for the GHG fees case. Fuel prices for each scenario are shown in Table 1 (including the GHG fee adders). Two natural gas prices are used for each case, depending on geographic location. Liquid fuel prices are the same for the entire system, but coal prices vary by state and some states have several different prices. GHG fees are added based on the heat input and type of fuel used by each unit. Fees imposed account for direct combustion emissions of CO2 as well as upstream emissions of CO2 and methane. Emissions rates and fees are detailed in Appendix B.
4. Results and discussion 3.3. Emissions modeling 4.1. Electricity generation mix To estimate emissions for each scenario at hourly resolution, dispatch model results for electricity generation were combined with emissions factors estimated from unit-specific historical data, with adjustments for anticipated emissions controls. The most detailed analysis was done for NOx emissions, due to the importance of timing and location to the air quality implications of this pollutant. Hourly emissions data for 2011 for fossil fuel power plants in the western U.S. were downloaded from the EPA Air Markets Program Data website (ftp:// ftp.epa.gov/dmdnload/emissions/smoke/). The year 2011 was chosen to match the historical base year we plan to use for air quality modeling in future research. For each unit, the data include hourly emissions of NOx and SO2 in units of pounds per hour, hourly heat input in units of million BTU (MMBtu) and gross load in units of MW. (Original units are retained here to match the convention in the U.S. air pollution control field. Conversions to standard SI units are as follows: 1 Btu = 1.055 kJ; 1 lb = 0.454 kg; 1 lb/MMBtu = 0.430 g/MJ.) Emissions-to-load relationships were studied for each individual unit in Colorado, New Mexico, Utah and Wyoming. For power plants outside of those four states, we estimated these relationships individually for the largest power plants, accounting for 70% of the generation. For the remaining power plants, we applied average emissions factors by plant type – either coal, natural gas combined cycle, or natural gas combustion turbine. As illustrated for NOx emissions in Appendix C Figs. C1 and C2, for individual units the relationship between NOx or SO2 emissions and power plant gross load is generally linear (but with a non-zero intercept) when the load is higher than 40% of maximum gross load in the case of coal power plants and above 50% of the maximum gross load in the case of natural gas power plants. Especially for natural gas power plants, however, NOx emissions are relatively high per unit generation at low load, compared to stabilized emissions rates. To account for this effect, we estimated separate NOx emissions versus load relationships
For each scenario, a yearlong simulation was performed with daily steps of 24 h-long intervals, which included hourly regional level load data and unit level renewable generation data. Fig. 2 shows the total annual generation broken down by type for each scenario. As shown, the natural gas prices, greenhouse gas fees and availability of renewables clearly affect the generation mix. As expected, natural gas generation is highest in the cheap gas scenario, accounting for 22% of generation. Natural gas accounts for 12% of generation in the GHG fees scenario (which has significantly more wind generation compared to
Fig. 2. Annual generation by type for four different scenarios within the Western Interconnection.
604
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
Fig. 4. Time series generation dispatch stack for all of the Western Interconnection for the first week of July (typical summer week). Fig. 3. Annual generation by type for geographic sub-regions within the Western Interconnection.
the other scenarios, based on the MARKAL model outputs for wind capacity). Coal provides about 25% of generation in the base case, declining to 17% in the GHG fees case. Modest curtailment of renewables is seen in the GHG fees scenario, which has the highest wind capacity. Fig. 3 shows the breakdown of annual generation for sub-regions within the Western Interconnection area for the cheap gas, costly gas, and GHG fees cases. (The base case is omitted because results are similar to those for the cheap gas case.) Similar to what was shown in Fig. 2, in Fig. 3 the largest differences between the scenarios are the changes in wind generation, specifically in CA, CO + WY, and WA + OR + ID, and the reductions in natural gas and coal generation in AZ, CA, CO + WY and WA + OR + ID. Total generation increases in CO + WY in the GHG fees scenario, as wind power is exported to other areas. There is also increased curtailment in CO + WY and CA in the GHG fees case, which suggests that either the system cannot fully utilize the available generation, there is no available transmission to export it, or both. Committed fossil fuel generator minimum generation levels can limit the amount of local renewable generation that can be used since they cannot be decommitted. If nearby transmission lines are at their flow limits or the areas where the power needs to be exported are already congested, that renewable generation must be curtailed. Air quality implications of electricity generation are not just determined by the annual generation mix, but also by the generation mix over shorter time scales. To illustrate the short-term variations, Fig. 4 shows an hourly dispatch stack for the entire Western Interconnection for the first week in July for the cheap gas, costly gas, and GHG fees scenarios. (Again the base case is excluded because results are similar to those for the cheap gas case.) This is a typical summer time period, with relatively high load and lower wind resources than other times of year. During this time period, coal accounts for 22% of generation in the cheap gas case, 23% in the costly gas case, and 21% in the GHG fees case. The contribution from natural gas is 25% in the cheap gas case, 17% in the costly gas case, and 16% in the GHG fees case. Towards the end of the week the wind resource increases, decreasing the amount of
Fig. 5. Time series generation dispatch stack for the Colorado and Wyoming sub-region for the first week of July (typical summer week). Dashed line shows load in the subregion.
natural gas and coal used in the costly gas and GHG fees cases. In the GHG fees case, wind generation displaces coal through more frequent and larger magnitude ramping. Fig. 5 shows another level of detail for the hourly dispatch in July, focusing in on the CO + WY region (Public Service Company of Colorado (PSCO), PacifiCorp East – Wyoming (PAWY), and Western Area Power Administration, Colorado-Missouri Region (WACM) balancing
605
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
Table 2 Annual emissions from coal and natural gas fired generating units for the U.S. portion of the Western Interconnection. Percent reductions from the baseline case are shown in parentheses. Scenario
CO2 from Coal (million metric tons)
CO2 from NG (million metric tons)
Combined CO2 (million metric tons)
Baseline Cheap Gas Costly Gas GHG Fees
206 192 (7%) 173 (16%) 134 (35%)
55 68 (−24%) 32 (42%) 28 (49%)
261 260 (0.4%) 205 (21%) 162 (38%)
Scenario
SO2 from Coal (thousand metric tons)
SO2 from NG (thousand metric tons)
Combined SO2 (thousand metric tons)
Baseline Cheap Gas Costly Gas GHG Fees
114 104 (9%) 94 (18%) 71 (38%)
0.3 0.4(−33%) 0.2(33%) 0.2(33%)
114 104 (9%) 94 (18%) 71 (38%)
Scenario
NOx from Coal (thousand metric tons)
NOx from NG (thousand metric tons)
Combined NOx (thousand metric tons)
Baseline Cheap Gas Costly Gas GHG Fees
183 167 (9%) 149 (19%) 111 (39%)
12 16 (−33%) 7 (42%) 6 (50%)
196 183 (7%) 156 (20%) 118 (40%)
Fig. 6. Hourly NOx emissions projections from coal and natural gas power plants in the U.S. portion of the Western Interconnection for the cheap gas, costly gas and GHG fees scenarios.
18% lower in this case than in the base case. Total SO2 emissions in the GHG fees scenario are about 38% those in the base scenario. Emissions of NOx are also dominated by coal-fired generation, with emissions from natural gas-fired generation accounting for less than 10% of the total in all scenarios. Compared to the base case, total annual NOx emissions are reduced by about 7% in the cheap gas case, 20% in the costly gas case, and 40% in the GHG fees case. However, as shown below, the differences between the scenarios vary significantly over the year. Fig. 6 shows the area-wide hourly NOx emissions over the year for the cheap gas, costly gas, and GHG fee scenarios. (Again, results for the base case are similar to those for the cheap gas case.) For coal, while the GHG fee scenario clearly has lower emissions than other scenarios, the differences are most pronounced in spring and fall. During the summer season, when NOx emissions contribute most to ozone formation, the combined emissions from coal-and natural gas-fired power plants are similar across all four scenarios. In the July to September period, electricity demand increases and availability of wind decreases so that many coal and gas-fired power plants must be dispatched, regardless of scenario. As expected, NOx emissions from natural gas power plants are highest in the cheap gas scenario. These emissions show a pronounced summer peak, again corresponding to high load and relatively low wind resource availability. In particular, while generation from natural gas combined cycle plants occurs year-round in all the scenarios, operation of less efficient and higher-emitting combustion turbines is mainly limited to the summer peak period. As noted above, compared to the base case, total annual power plant NOx emissions across the region are reduced by about 7% in the cheap gas case, 20% in the costly gas case, and 40% in the GHG fees case. However, the differences across cases are sharply curtailed if only the summer months are considered. For example, averaged over July and August, power plant NOx emissions in the cheap gas, costly gas, and GHG fees cases are only 2%, 4%, and 8% lower than those in the base case. This dampened effect is shown in more detail in Fig. 7, which presents total NOx emissions from coal and natural gas-fired electricity generation for the first week of July. Results for the U.S. portion of the Western Interconnection are shown in Fig. 7a; results for CO + WY in Fig. 7b. During this time period, hourly emissions rates for the cheap and costly gas cases cross, depending on the day and time. This occurs for both the Western Interconnection domain and the CO + WY subregion. On several days during this period, NOx emissions from natural gas plants in the costly gas case are about equal to those in the GHG fees
areas). This area shows larger differences between scenarios than the plot for the entire interconnection. As expected, the cheap gas case has the most natural gas generation. In the costly gas case, wind generation is much larger and almost completely displaces natural gas except for the evening peak load hours and when wind generation decreases. Coal generation is similar to that in the cheap gas case, although there is an increase in ramping frequency. The GHG fees case has more wind generation and less coal generation than any of the other scenarios. Similar to the costly gas case, there is little natural gas generation. Coal generation ramping is increased in frequency, ramp rate, and overall magnitude and appears to be independent of load. 4.2. Emissions Table 2 shows the consequences of each of the scenarios for annual SO2, CO2 and NOx emissions from electricity generation in the Western Interconnection region, excluding portions in Mexico and Canada. Emissions are impacted by both long-term decisions (the type of generation that is built) and short-term decisions (how different generators are operated). In the cheap gas case, emissions reductions relative to the base case are primarily due to running gas generators more than coal. In the costly gas and GHG fees scenarios, emissions reductions are primarily due to increased penetration of renewables. As expected, the GHG fees scenario has the lowest CO2 and SO2 emissions. Total CO2 emissions are similar in the base and cheap gas cases, as increased emissions from natural gas generation offset reductions in emissions from coal. Total CO2 emissions are about 20% lower in the costly gas case than in the base and cheap gas cases; increased renewable penetrations lead to sharp reductions in natural gas use combined with more modest reductions in generation from coal. Total CO2 emissions in the GHG fees case are about 38% lower than those in the base and cheap gas cases due to renewable generation displacing both gas and coal. SO2 emissions in all cases are dominated by coal-fired generation, as other resources have low or no emissions. Correspondingly, SO2 emissions are about 9% lower in the cheap gas case than in the base case, as less coal is used in the cheap gas case. Because wind displaces some coal as well as natural gas in the costly gas case, SO2 emissions are about 606
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
corresponding air pollutant emissions in the western U.S., circa 2030. In contrast to typical emissions scenario analyses that consider annual and regionally averaged results, this study focuses on understanding how emissions responses might differ within the region and over the year. Although averaged results are adequate for CO2 emissions, higher spatial and temporal resolution is needed to understand local air quality implications for SO2 and NOx emissions. We focused here on NOx emissions due to the contribution this pollutant makes to elevated ozone concentrations in the western U.S. The results of our study show the potential for significant shifts in the electricity generation mix across the region, depending on fuel prices and whether or not emissions fees are applied. Fuel prices and fees specified for each scenario in the prior study with MARKAL (McLeod et al., 2014) determine the renewable energy capacity assumed for the region, which is input to the dispatch model. Across the region, electricity from wind ranges from less than 10% of generation in the base and cheap gas cases to 26% in the GHG fees case. Solar energy contributes about 7% of generation in all of the scenarios. With the renewables levels given, fuel prices and fees further influence the dispatch split between natural gas and coal. Natural gas contributes from 12% of annual generation in the region in the GHG fees case to 22% of generation in the cheap gas case. Coal-fired power plants contribute from 17% of generation in the GHG fees case to 25% in the base case. The generation mix and scenario responses show pronounced differences across the Western Interconnection region. Natural gas generation occurs mainly in CA and the WA + OR + ID sub-region, especially in the base (not shown) and cheap gas scenarios. Coal generation is used most heavily in AZ and CO + WY, in the base, cheap, and costly gas scenarios. Wind displaces coal in CO + WY, with wind generation increasing for export to other sub-regions in the costly gas and especially the GHG fees cases. Generation shifts between wind and natural gas on the West Coast. Responses in the four scenarios also vary sharply across the year. In particular, during summer, relatively full utilization of coal and natural gas plants is required to meet high demand, regardless of fuel prices and emissions fees. Annual results for the U.S. portion of the Western Interconnection show reductions in CO2 emissions of < 1%, 20% and 38% in the cheap gas, costly gas, and GHG fees cases, respectively, compared to the base case. Region-wide annual reductions in SO2 emissions are 9%, 18% and 38% in the cheap gas, costly gas, and GHG fees cases, respectively; reductions in NOx emissions are 7%, 20%, and 40%. Ignoring the difference in start-up versus stable NOx emissions rates for natural gas plants has little impact on region-wide emissions estimates, but would be more important in the local vicinity of some of these units. A key insight from downscaling the emissions estimates is that NOx emissions rates are highest in summer in all scenarios, due to relatively heavy reliance on coal and natural gas in that season. The seasonal enhancement is greatest for the GHG fees case, in which reductions compared to other cases occur in spring, fall, and winter, but not in the summer. In that case, across the U.S. portion of the Western Interconnect, the average NOx emissions rate in July and August is more than 60% higher than the average emissions rate for the year. Furthermore, during the summer period the NOx emissions reductions that are seen occur mainly at night. Accounting for seasonal and diurnal timing is important, because NOx emissions reductions are most effective at reducing ozone pollution if they occur on hot, stagnant summer days when meteorological conditions are most conducive to the formation and build-up of ozone from photochemical processes (NRC, 1991; Mesbah et al., 2015). Our study thus suggests that ozone pollution benefits might be overestimated if policy analyses rely on emissions reduction estimates with only annual time resolution. Emissions reductions that occur outside the summer season are still beneficial, however, because NOx emissions also contribute to regional haze, to health effects from secondary aerosols, and to nitrogen deposition that harms high altitude ecosystems in the western U.S. Results of this scenario analysis are not intended to be taken as
Fig. 7. Hourly total NOx emissions (coal and natural gas combined) in (a) the U.S. portion of the Western Interconnection and (b) the CO + WY sub-region for a typical summer week.
case and are less than half those in the cheap gas case (separate results not shown). However, because coal-fired power plant emissions dominate overall, the difference in natural gas plant emissions is offset by coal-plant emissions that are modestly lower in the cheap gas case. Again considering either the whole region or the CO + WY sub-region, NOx emissions reductions in the GHG fees case during this week occur mainly at night. In contrast, reductions in afternoon peak period emissions are comparatively modest. Results shown for NOx emissions in Table 2 and in Figs. 6 and 7 were computed using differentiated emissions versus load relationships for start-up conditions. The effect of doing so is shown in Appendix C, considering both annual and July/August averaging periods. Overall, when the calculations are simplified by using average relationships encompassing start-up periods together with stabilized loads, total NOx emissions are underestimated by less than 1% in all cases. However, NOx emissions from natural gas-fired power plants are underestimated by up to 14%, which occurs for the GHG fees case in July/August. 5. Conclusions and policy implications This study examines how future scenarios for natural gas prices and GHG emissions fees might influence the electricity generation mix and 607
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
energy source, Energy Information Administration, 〈https://www.eia.gov/ electricity/data/state/〉 (Accessed 14 July 2016). EIA, 2016e. How much carbon dioxide is produced per kilowatthour when generating electricity with fossil fuels? Energy Information Administration. 〈www.eia.gov/ tools/faqs/faq.cfm?Id=74&t=11〉 (Accessed 16 January 2017). EIA, 2016f. Annual energy outlook 2016: with projections to 2040, Energy Information Administration, 〈https://www.eia.gov/outlooks/archive/aeo16/〉 (Accessed 9 February 2017). EPA, 2014. Regulatory impact analysis for the proposed carbon pollution guidelines for existing power plants and emission standards for modified and reconstructed power plant. U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards, Research Triangle Park, NC. EPA, 2016. Cross-state air pollution rule update for the 2008 ozone NAAQS, 81 Fed. Reg. 74504, October 26. EPA, 2016a. 2015 National Ambient Air Quality Standards (NAAQS) for Ozone. U.S. Environmental Protection Agency. 〈https://www.epa.gov/ozone-pollution/2015national-ambient-air-quality-standards-naaqs-ozone〉 (Accessed 14 July 2016). EPA, 2016b. Air markets program data, U.S. Environmental Protection Agency. 〈https:// ampd.epa.gov〉 (Accessed 16 January 2017). Eurek, K., Cole, W., Bielen, D., Blair, N., Cohen, S., Frew, B., Ho, J., Krishnan, V., Mai, T., Sigrin, B., Steinberg, D., 2016. Regional Energy Deployment System (ReEDS) Model Documentation: Version 2016, NREL/TP-6A20-67067. National Renewable Energy Laboratory, Golden, CO. Gilbraith, N., Powers, S.E., 2013. Residential demand response reduces air pollutant emissions on peak electricity demand days in New York City. Energy Policy 59, 459–469. http://dx.doi.org/10.1016/j.enpol.2013.03.056. Hobbs, B., Hu, M.-C., Chen, Y., Ellis, J.H., Paul, A., Burtraw, D., Palmer, K.L., 2010. From regions to stacks: spatial and temporal downscaling of power pollution scenarios. IEEE Trans. Power Syst. 25 (2), 1179–1189. IWG, 2013. Technical support document: technical update of the social cost of carbon for regulatory impact analysis, Interagency Working Group on Social Cost of Carbon, Washington, DC. Kerl, P.Y., Zhang, W., Moreno-Cruz, J.B., Nenes, A., Realff, M.J., Russell, A.G., Sokol, J., Thomas, V.M., 2015. New approach for optimal electricity planning and dispatching with hourly time-scale air quality and health considerations. Proc. Natl. Acad. Sci. USA 201413143. Krieger, E.M., Casey, J., Shonkoff, S.B.C., 2016. A framework for siting and dispatch of emerging energy resources to realize environmental and health benefits: case study on peaker power plant displacement. Energy Policy 96, 302–313. Lenox, C.D., Gage, C., Kaplan, O., Loughlin, D., Yelverton, W., 2013. EPA U.S. Nine-region MARKAL database: Database documentation. Environmental Protection Agency, Cincinnati, OH, U.S. Martinez-Anido, C.B., Botor, B., Florita, A.R., Draxl, C., Lu, S., Hamann, H., Hodge, B., 2016. The value of day-ahead solar power forecasting improvement. Sol. Energy 129, 192–203. McLeod, J., Brinkman, G., Milford, J.B., 2014. Emissions implications of future natural gas production and use in the U.S. and in the Rocky Mountain region. Environ. Sci. Technol. 48, 13036–13044. Mesbah, S.M., Hakami, A., Schott, S., 2015. Optimal ozone control with inclusion of spatiotemporal marginal damages and electricity demand. Environ. Sci. Technol. 2015 (49), 7870–7878. http://dx.doi.org/10.1021/acs.est.5b01178. Novacheck, J., Johnson, J., 2017. Diversifying wind power in real power systems. Renew. Energy 106, 177–185. NRC, 1991. Rethinking the Ozone Problem in Urban and Regional Air Pollution. National Research Council, Committee on Tropospheric Ozone. National Academy Press, Washington, DC. Pacsi, A., Alhajeri, N.S., Zavala-Araiza, D., Webster, M.D., Allen, D.T., 2013. Regional air quality impacts of increased natural gas production and use in Texas. Environ. Sci. Technol. 47 (7), 3521–3527. http://dx.doi.org/10.1021/es3044714. Pacsi, A., Kimura, Y., McGaughey, G., McDonald-Buller, E.C., Allen, D.T., 2015. Regional ozone impacts of increased natural gas use in the texas power sector and development in the Eagle Ford Shale. Environ. Sci. Technol. 49, 3966–3973. http://dx.doi.org/10. 1021/es5055012. Plachinski, S.D., Holloway, T., Meier, P.J., Nemet, G.F., Rrushaj, A., Oberman, J.T., Duran, P.L., Voight, C.L., 2014. Quantifying the emissions and air quality co-benefits of lower-carbon electricity production. Atmos. Environ. 94, 180–191. Thompson, T.M., King, C.W., Allen, D.T., Webber, M.E., 2011. Air quality impacts of plugin hybrid electric vehicles in Texas: evaluating three battery charging scenarios. Environ. Res. Lett. 6, 024004. Thompson, T.M., Rausch, S., Saari, R.K., Selin, N.E., 2014. A systems approach to evaluating the air quality co-benefits of US carbon policies. Nat. Clim. Change 4, 917–923. Thompson, T., Rausch, S., Saari, R.K., Selin, N.E., 2016. Air quality co-benefits of subnational carbon policies. J. Air Waste Manag. Assoc. http://dx.doi.org/10.1080/ 10962247.2016.1192071. Trail, M., Tsimpidi, A.P., Liu, P., Tsigaridis, K., Rudokas, J., Miller, P., Nenes, A., Hu, Y., Russell, A.G., 2014. Sensitivity of air quality to potential future climate change and emissions in the United States and major cities. Atmos. Environ. 94, 552–563. WECC, 2015. TEPPC study report – 2024 PC1 common case and release notes for WECC 2024 common case version 1.1, Western Electricity Coordinating Council, 〈https:// www.wecc.biz〉.
projections of future emissions. Rather, the study is meant to provide insight into potential trends and tradeoffs between contrasting scenarios, while recognizing limitations in their assumptions. These limitations include the use of a single historical year for meteorological inputs and the load profile, which provides an inherently limited representation of what future conditions might be like. We used the latest information on utilities’ plans that was available at the time this study was started, but those plans continue to evolve. The dispatch results are constrained by the electricity transmission infrastructure assumed in the model; future changes to this system would alter the results. Furthermore, fuel price forecasts are uncertain and evolving; more recent projections suggest future natural gas prices could fall below those assumed in our cheap gas scenario (EIA, 2016f). The air quality implications of the four electricity generation scenarios are being investigated in ongoing research that will use the detailed emissions results from this study in an atmospheric chemistry and transport model. Funding Support for this work was provided by the National Science Foundation (NSF) AirWaterGas Sustainability Research Network CBET1240584. Findings or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the NSF. Acknowledgments The authors thank Kent Kurashima for assistance with data analysis and graphics and Shannon Capps for feedback on early drafts of the results. Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.enpol.2017.07.051. References Bloom, A., Townsend, A., Palchak, D., Novacheck, J., King, J., Barrows, C., Ibanez, E., O’Connell, M., Jordan, G., Roberts, B., Draxl, C., Gruchalla, K., 2016. Eastern Renewable Generation Integration Study, NREL/TP-6A20-64472. National Renewable Energy Laboratory, Golden, CO. Brinkman, G.L., Denholm, P., Hannigan, M.P., Milford, J.B., 2010. Effects of plug-in hybrid electric vehicles on ozone concentrations in Colorado. Environ. Sci. Technol. 44, 6256–6262. Brinkman, G., Jorgenson, J., Ehlen, A., Caldwell, J., 2016. Low Carbon Grid Study: Analysis of a 50% Emission Reduction in California, NREL/TP-6A20-64884. National Renewable Energy Laboratory, Golden, CO. Brown, K.E., Henze, D.K., Milford, J.B., 2013. Accounting for climate and air quality damages in future US electricity generation scenarios. Environ. Sci. Technol. 47 (7), 3065–3072. Brown, K.E., Henze, D.K., Milford, J.B., 2017. How accounting for climate and health impacts of emissions could change the US energy system. Energy Policy 102, 396–405. Buonocore, J.J., Luckow, P., Norris, G., Spengler, J.D., Biewald, B., Fisher, J., Levy, J.I., 2015. Health and climate benefits of different energy-efficiency and renewable energy choices. Nat. Clim. Change 6, 100–107. http://dx.doi.org/10.1038/ NCLIMATE2771. Cole, W., Mai, T., Logan, J., Steinberg, D., McCall, J., Richards, J., Sigrin, B., Porro, G., 2016. 2016 Standard scenarios report: a U.S. electricity sector outlook, National Renewable Energy Laboratory Technical Report NREL/TP-6A20-66939, Golden, CO. EIA, 2013. Annual energy outlook 2013. Energy Information Administration, Washington, DC. EIA, 2016a. State energy data system, Energy Information Administration. 〈https://www. eia.gov/state/?Sid=US〉 (Accessed 14 July 2016). EIA, 2016b. Monthly energy review: coal, Energy Information Administration., 〈https:// www.eia.gov/totalenergy/data/monthly/〉 (Accessed 14 July 2016). EIA, 2016c. Natural gas data: natural gas withdrawals and production, Energy Information Administration, 〈https://www.eia.gov/naturalgas/data.cfm〉 (Accessed 14 July 2016). EIA, 2016d. Electricity, detailed state data, net generation by state by type of producer by
608
Contents lists available at ScienceDirect
Energy Policy journal homepage: www.elsevier.com/locate/enpol
Emissions implications of downscaled electricity generation scenarios for the western United States Rene Nsanzinezaa, Matthew O’Connellb, Gregory Brinkmanb, Jana B. Milforda, a b
MARK
⁎
Department of Mechanical Engineering, University of Colorado, 1111 Engineering Drive, Boulder, CO 80309-0427, USA Strategic Energy Analysis Center, National Renewable Energy Laboratory, 15013 Denver West Parkway, Golden, CO 80401, USA
A R T I C L E I N F O
A B S T R A C T
Keywords: Electricity dispatch Air quality Emissions fees Nitrogen oxides Natural gas Renewable energy
This study explores how emissions from electricity generation in the Western Interconnection region of the U.S. might respond in circa 2030 to contrasting scenarios for fuel prices and greenhouse gas (GHG) emissions fees. We examine spatial and temporal variations in generation mix across the region and year using the PLEXOS unit commitment and dispatch model with a production cost model database adapted from the Western Electricity Coordinating Council. Emissions estimates are computed by combining the dispatch model results with unitspecific, emissions-load relationships. Wind energy displaces natural gas and coal in scenarios with relatively expensive natural gas or with GHG fees. Correspondingly, annual emissions of NOx, SO2, and CO2 are reduced by 20–40% in these cases. NOx emissions, which are a concern as a precursor of ground-level ozone, are relatively high and consistent across scenarios during summer, when peak electricity loads occur and wind resources in the region are comparatively weak. Accounting for the difference in start-up versus stabilized NOx emissions rates for natural gas plants had little impact on region-wide emissions estimates due to the dominant contribution from coal-fired plants, but would be more important in the vicinity of the natural gas units.
1. Introduction Led by Wyoming, the eight states in the Rocky Mountain region (AZ, CO, ID, MT, NM, NV, UT, WY) account for more than 50% of U.S. coal production and 18% of natural gas production (EIA, 2016a). Coal production in the region peaked at over 600 million short tons in 2008 and fell to 480 million tons in 2015 (EIA, 2016a, 2016b). In 2015, the states in the Rocky Mountain region produced 5.1 trillion cubic feet of natural gas, up 6% from the level in 2005, with sharp growth in Colorado offsetting a decline in New Mexico and relatively stable production in Wyoming (EIA, 2016a, 2016c). Since 2010, utilities in the region have completed or announced plans to retire or repower more than 20 coal-fired power plant units. Electricity generation in these states shifted from 63% coal, 19% natural gas and less than 1% wind in 2005, to 52% coal, 22% natural gas and 5% wind in 2014 (EIA, 2016d). Shifts to natural gas and renewable energy are expected to continue in the western U.S., with significant implications for the region's air quality. Electricity generation and energy production activities are
major sources of nitrogen oxides (NOx), sulfur oxides, volatile organic compounds (VOC) and greenhouse gases (GHG). Except for Idaho and Montana, all of the states in the Rocky Mountain region include areas that have been designated nonattainment for the 2008 National Ambient Air Quality Standard for ozone or have recent ozone values above the 2015 revised standard (EPA, 2016a), so emissions of ozone precursors – NOx and VOCs – are of particular concern. Understanding how the changing energy landscape might affect future emissions and air quality in the Rocky Mountain region requires a combination of energy and air quality models that bridge across hourly to decadal time scales. The objective of this study is to examine the sub-regional spatial patterns and sub-annual temporal patterns of electricity sector emissions within the Rocky Mountain region that might result from contrasting scenarios for future natural gas prices and greenhouse gas mitigation policies. The study builds on prior work by McLeod et al. (2014), who used the U.S. Environmental Protection Agency's (EPA) nine-region MARKAL energy system model to examine how annual
Abbreviations: CSP, Concentrating Solar Power; EIA, Energy Information Administration; EPA, Environmental Protection Agency; ERCOT, Electric Reliability Council of Texas; GHG, greenhouse gases; NREL, National Renewable Energy Laboratory; PAWY, PacifiCorp East – Wyoming; PSCO, Public Service Company of Colorado; PV, photovoltaics; RPV, rooftop photovoltaics; TEPPC, Transmission Expansion Planning Policy Committee; VOC, volatile organic compounds; WACM, Western Area Power Administration – Colorado-Missouri; WECC, Western Electricity Coordinating Council ⁎ Corresponding author. E-mail addresses: [email protected] (R. Nsanzineza), [email protected] (M. O’Connell), [email protected] (G. Brinkman), [email protected] (J.B. Milford). http://dx.doi.org/10.1016/j.enpol.2017.07.051 Received 20 February 2017; Received in revised form 6 June 2017; Accepted 23 July 2017 0301-4215/ © 2017 Published by Elsevier Ltd.
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
meteorological conditions are conducive to ozone formation. Fig. A1 in Appendix A illustrates such conditions, showing the correlation between temperature, power plant load and ozone for Denver, CO in July 2006. Recent studies have started to conduct electricity system analyses that provide higher resolution, by using electricity dispatch models or simplified representations of dispatch order to estimate emissions changes in response to altered demand, fuel prices, or policies to encourage renewables (e.g., Hobbs et al., 2010; Brinkman et al., 2010; Thompson et al., 2011; Plachinski et al., 2014). For example, Gilbraith and Powers (2013) applied a dispatch model to simulate residential demand response in New York City, finding that a moderate program could reduce generation from small, relatively inefficient combustion turbines used to meet peak demand, thus reducing NOx and PM2.5 emissions on poor air quality days. Buonocore et al. (2015) applied a dispatch model for the Eastern Interconnection to compare the emissions and corresponding health benefits of incorporating wind, solar, and demand response at six locations in the eastern U.S., illustrating how the benefits depend on the type and location of fossil fuel generation being displaced. Kerl et al. (2015) used a reduced-form air quality model relating emissions to pollutant concentrations and monetized estimates of resulting health effects to generate environmental damage costs for inclusion in a least-cost electricity dispatch algorithm for power plants in Georgia. Including environmental costs in the dispatch algorithm led to shifting generation on some winter days from coal-fired power plants in northern Georgia to a natural gas combined cycle plant near the coast. Pacsi et al. (2013) used the PowerWorld Simulator to estimate how hourly generation from units in the existing Electric Reliability Council of Texas (ERCOT) system depends on natural gas prices, as an input to an assessment of net air quality impacts from natural gas production in the Barnett shale. They estimated NOx emissions for each unit by multiplying hourly generation with the unit's annual average emissions factor. Net impacts on ozone air quality in Texas were modeled using a regional atmospheric chemistry and transport model, CAMx (Comprehensive Air Quality Model with Extensions), adjusting the power plant NOx emissions and estimates of NOx and VOC emissions from local natural gas production in each gas price case. Pacsi et al. (2015) conducted a similar net air quality analysis assuming the natural gas came from the Eagle Ford shale, instead of the Barnett. Compared to prior emissions studies using dispatch modeling, this study is unique in focusing on the Rocky Mountain region, which has abundant wind and solar resources as well as coal and natural gas. We examine relatively broad and self-consistent scenarios for key factors that could shape future electricity generation in the region, covering a range of natural gas prices and greenhouse gas mitigation policies. Unlike other studies that have considered limited adjustments to the current electricity system, we examine dispatch results for a future electricity generating fleet that reflects utilities’ current plans for retiring or repowering coal plants. Lastly, unlike most prior studies, we use detailed unit-specific emissions models that account for load-dependence of emissions rates.
average emissions would respond to these scenarios out to the year 2050, for the entire U.S. and for the Rocky Mountain region as a whole. To begin to examine spatial and temporal patterns within the Rocky Mountain region, the current study “downscales” the electricity sector emissions estimates for this region for a circa 2030 time period. To do this, we simulate the electricity generation mix in selected scenarios at hourly temporal resolution using the PLEXOS unit commitment and dispatch model for the power plant fleet expected to be in place in the Western Interconnection at that time. Hourly variations in hydro, wind and solar resource availability and electricity demand are estimated from historical data, which are then scaled by increased overall demand and renewable electricity capacity for the future scenarios. Emissions are computed by combining unit-specific loads from the dispatch model with load-dependent emissions factors. To our knowledge, this is the first such study for the Rocky Mountain region, and the first study in any U.S. region to provide this level of detail for a future timeframe when significantly expanded capacity for renewable energy could be in place to compete with natural gas and coal. Future work will combine the electricity sector emissions results described here with estimates of emissions changes from upstream energy production, and use these as inputs to a regional-scale chemistry and transport model to study net air quality impacts of natural gas production and use. 2. Background The U.S. EPA, U.S. Energy Information Administration (EIA) and other Department of Energy laboratories have long used least-cost or partial equilibrium energy system or power sector planning models to examine emissions impacts of future electricity generation or broader energy scenarios and to analyze proposed emissions control strategies and other regulations. These models include EIA's National Energy Modeling system (EIA, 2013) used to produce their Annual Energy Outlook; EPA's MARKAL model and nine-region U.S. database (Lenox et al., 2013); the Integrated Planning Model (EPA, 2014) used in regulatory analyses such as those for EPA's Cross-State Air Pollution Rule (EPA, 2016); and the National Renewable Energy Laboratory's Regional Energy System Deployment (ReEDS) model (Eurek et al., 2016; Cole et al., 2016). Researchers have used these and similar models to examine a range of scenarios and policy proposals. For example, Brown et al. (2013, 2017) applied MARKAL with EPA's nine-region database to examine how adding damage-based fees in the electric power and other U.S. energy sectors would alter emissions. Trail et al. (2014) used emissions from EPA's MARKAL reference case to examine how air quality across the U.S. might change by 2050. Thompson et al. (2014, 2016) used the U.S. Regional Energy Policy model to estimate conventional air pollutant emissions responses to greenhouse gas mitigation strategies. In work that forms the basis for the current study, McLeod et al. (2014) modified EPA's MARKAL model with updated information on costs of renewable energy and emissions from oil and gas production, then used the revised model to examine how contrasting scenarios for natural gas supply and demand, constraints on the electricity generation mix, and GHG fees might affect energy system emissions for the U.S. and the Rocky Mountain region out to the year 2050. These studies all focus on projecting annual average emissions changes at the state, regional, or national scale. In studies that estimate air quality impacts from the energy and emissions modeling results, the projected annual changes in emissions are typically used to scale existing emissions inventories up or down, assuming the location and subannual timing of emissions will not change (e.g., Thompson et al., 2014, 2016). Higher resolution in emissions projections is critical for examining how alternative economic or policy scenarios might influence emissions during specific time periods that are prone to air quality problems (Hobbs et al., 2010; Krieger et al., 2016), e.g., on hot summer days when electricity demand is high, wind availability tends to be low, and
3. Methods 3.1. Scenarios The natural gas price and emissions fee scenarios and corresponding renewable energy build outs considered in this study were developed by McLeod et al. (2014), using the MARKAL model with the EPA U.S. nineregion database. The MARKAL model finds the least cost means to satisfy future end use demand in the industrial, commercial, residential, and transportation sectors, under specified constraints including limits on fuel supplies and on rates of capacity expansion and introduction of new technology. The EPA database available at the time was developed to match projections from Energy Information Administration's 2012 Annual Energy Outlook. McLeod et al. (2014) modified the 2012 602
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
Anido, 2016; Novacheck and Johnson, 2017). What this means is that PLEXOS will simulate the unit commitment (decisions to start up or shut down) and dispatch (adjusting generation levels) of electrical generation units to match grid energy demand for the lowest cost possible solution, subject to a suite of constraints. These constraints include transmission line flow limits, minimum and maximum generation levels, generation ramping limits, and maximum or minimum energy output per month for hydro plants. Costs include fuel cost, start and shutdown costs, variable operation and maintenance costs, and emissions fees (if applicable). The PLEXOS software was run at the National Renewable Energy Laboratory (NREL) using a modified version of the Western Electricity Coordinating Council's (WECC) Transmission Expansion Planning Policy Committee's (TEPPC) 2024 Common Case production cost model database (WECC, 2015). This database represents the expected 2024 Western Interconnection electrical grid and generation fleet. We made minor changes to the 2024 TEPPC database in order to implement the desired scenarios. The database contains 4836 generators and 43 individual load regions, with a total of 8349 transformers, 16,847 transmission lines, and 19,782 individual nodes. The generators have unique properties specified such as fuel type, heat rate, maximum capacity, minimum generation levels, operating cost, ramp rate limitations, and minimum run time and down time. Hourly load profiles are based on the actual hourly loads from 2006, scaled to expected 2024 load magnitude. Hourly load profiles are provided in the TEPPC 2024 Common Case database and estimate demand in 2024 for each balancing area, with a total interconnection peak demand of 178.9 GW occurring on July 24th at 5:00 P.M. PST. For purposes of this study, we assume the 2024 fleet configuration and demand estimates are representative for a circa 2030 timeframe. We used the TEPPC database as the best information available, and use the phrase “circa 2030” as the midpoint between the TEPPC load and fossil fleet infrastructure projections and the 2035 MARKAL results for renewables build out. The generation capacity of the database, broken down into individual areas and excluding variable generation renewables, is shown in Appendix B, Table B1. Variable generation technologies such as wind and solar PV are not shown because each case studied has a different renewable portfolio build out. As noted above, the renewable build outs are taken from McLeod et al. (2014). Note that use of the 2024 TEPPC database results in more coal capacity and coal generation than was found in the prior study with MARKAL, for which coal plant retirements in the region were further advanced. Renewable generation resources were developed for the model through a combination of existing projects, near-term projects under development, and estimated future projects. Generation profiles for these resources are based on 2006 meteorological conditions and were developed by NREL for TEPPC. Wind and solar generation in the PLEXOS simulations are based on these profiles, which are provided to the PLEXOS simulations as inputs. Hourly generation profiles are available for the entire year for all the wind and solar sites, and vary for each case. The database contains 417 wind generators and 596 solar generators, which are separated into three categories. There are 31 concentrated solar power (CSP) generators, 45 rooftop photovoltaic (RPV) sites representing many installations, and 520 utility scale photovoltaic (PV) installations. Total available generation from solar is held constant across scenarios at 78.6 TWh. Available generation from wind ranges from 82.7 TWh in the base case to 268 TWh in the GHG fees case. Appendix B Table B2 shows total generation from wind for each scenario, broken down by season. Solving an optimal power flow problem at the individual node level for the entire western interconnection is computationally difficult and would take an excessive amount of time. To decrease computing time, transmission networks within each region are aggregated into a single node, and all generators in the region are also assigned to this same node. Each region's load profile is represented 100% by its respective
version of EPA's nine-region energy system and emissions database to update cost and performance data for electric generating units, including for PV and wind, update renewable portfolio standards to reflect 2013 developments, separate unconventional and conventional natural gas production, and to account for emissions reductions in the oil and gas production sector due to EPA regulations promulgated in 2012. McLeod et al. (2014) modeled seven scenarios, four of which are considered as a starting point for this study. Results were examined at the national scale and specifically for the Rocky Mountain region. The base case reflects fuel supply and cost assumptions for the U.S. and Rocky Mountain region from the 2013 Annual Energy Outlook reference case. Two scenarios examined contrasting assumptions about future natural gas production, again based on AEO 2013, with “cheap gas” assuming relatively abundant supply and low prices, and “costly gas” assuming limited supply and high prices. The other scenario considered here used the base case assumptions for natural gas supplies, but applied GHG fees to CO2 and methane emissions in all sectors, including electricity generation and oil and gas production, based on the Social Cost of Carbon adopted by the Interagency Working Group in May 2013 (IWG, 2013). Fig. 1 is adapted from McLeod et al. (2014) and shows the electricity generation mix in the eight state Rocky Mountain region in the base year, 2010, and as modeled with MARKAL for 2035 in each of the four scenarios. (Results for the U.S. as a whole are given in McLeod et al., 2014.) For the Rocky Mountain region, electricity generation from natural gas and wind increases between 2010 and 2035. Wind displaces natural gas in the costly gas scenario, and displaces natural gas and coal in the scenario with greenhouse gas fees. Electricity generation in the costly gas and GHG fees scenarios is higher than in the base case due to increased electricity exports from the region. The current work uses estimates of renewable generation capacity, natural gas prices, and GHG fees from McLeod et al. (2014), but incorporates them into a unit commitment and dispatch model. Shifting to an hourly dispatch modeling framework provides the needed resolution to understand emissions implications for the electricity generation sector on the relatively short time-scale of air pollution episodes, and also allows consideration of how intraregional transmission constraints and ancillary service requirements might alter the choice between coal, natural gas, and renewables. 3.2. Dispatch modeling We used the PLEXOS unit commitment and economic dispatch model to simulate operation of the western US electric interconnection. PLEXOS is an optimization tool licensed and supported by Energy Exemplar that can solve unit commitment and economic dispatch problems (e.g., Bloom et al., 2016; Brinkman et al., 2016; Martinez-
Fig. 1. Rocky Mountain electricity generation mix, TWh (adapted from McLeod et al., 2014).
603
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
for stabilized (above 40% or 50% load) and low-load conditions. For comparison, we also estimate NOx emissions under the assumption that emissions are a constant multiplier of load across all load levels. Modeling emissions as a function of load and considering start-up emissions explains most of the additional emissions experienced during ramping and cycling operations. Further detail on the treatment of start-up emissions and discussion of their influence is provided in Appendix C. For power plants with pending control requirements that are tighter than those in place in 2011, we calculated an emissions control factor by comparing the maximum allowable rate in pending requirements to the maximum emissions rate in the 2011 dataset, using the averaging period specified in the pending requirements. If the maximum rate in 2011 was less than the pending limit, no adjustment was made to the historical emissions factors. For SO2 emissions, we used 2015 data from EPA's Air Markets Database (EPA, 2016b) to estimate unit specific emissions factors (tons/ MW gross load) for individual coal-fired power plants. These emissions factors were combined with annual electricity generation from the dispatch model results for each unit to estimate annual SO2 emissions. For natural gas power plants, we calculated average emissions factors by state and unit type (combustion turbine or combined cycle). The average emissions rate for natural gas by state was multiplied with the heat input data from the dispatch model to estimate the SO2 emissions from power plants in each of the future scenarios. Using statewide average emissions factors for natural gas power plants has little effect on the results because their SO2 emissions are low compared to those from coal-fired power plants. We estimated CO2 emissions for the future scenarios using emissions rates of 214 lb/MMBtu for coal and 117 lb/ MMBtu for natural gas (EIA, 2016e).
Table 1 Natural gas and coal fuel prices used in PLEXOS ($/MMBtu). Prices shown for the GHG fees case are added to those in the base case. Scenario
Natural gas in CA, OR, and WA
Natural gas elsewhere in Western Interconnection
Coal (price varies by state)
Base Cheap Gas Costly Gas GHG Fees
7.58 4.72 9.81 + 3.86
7.36 4.73 9.66 + 3.86
0.97 to 3.18 0.97 to 3.18 0.97 to 3.18 + 6.03
node. This simplification decreases computational time significantly, although due to the aggregation of lines it also reduces the accuracy of the solutions. In addition to this regional aggregation, all line reactance and resistances are removed and PLEXOS solves a transport flow problem, meaning the solver can decide how much power to send down each line to serve load in each corresponding region as long as line power flow constraints are not violated. The unit commitment and economic dispatch problem, including optimal power flow, was solved for the western interconnection for the four scenarios described previously. Natural gas prices used in PLEXOS are different in each scenario, but coal and liquid fuel (oil, propane, etc.) prices are the same except for the GHG fees case. Fuel prices for each scenario are shown in Table 1 (including the GHG fee adders). Two natural gas prices are used for each case, depending on geographic location. Liquid fuel prices are the same for the entire system, but coal prices vary by state and some states have several different prices. GHG fees are added based on the heat input and type of fuel used by each unit. Fees imposed account for direct combustion emissions of CO2 as well as upstream emissions of CO2 and methane. Emissions rates and fees are detailed in Appendix B.
4. Results and discussion 3.3. Emissions modeling 4.1. Electricity generation mix To estimate emissions for each scenario at hourly resolution, dispatch model results for electricity generation were combined with emissions factors estimated from unit-specific historical data, with adjustments for anticipated emissions controls. The most detailed analysis was done for NOx emissions, due to the importance of timing and location to the air quality implications of this pollutant. Hourly emissions data for 2011 for fossil fuel power plants in the western U.S. were downloaded from the EPA Air Markets Program Data website (ftp:// ftp.epa.gov/dmdnload/emissions/smoke/). The year 2011 was chosen to match the historical base year we plan to use for air quality modeling in future research. For each unit, the data include hourly emissions of NOx and SO2 in units of pounds per hour, hourly heat input in units of million BTU (MMBtu) and gross load in units of MW. (Original units are retained here to match the convention in the U.S. air pollution control field. Conversions to standard SI units are as follows: 1 Btu = 1.055 kJ; 1 lb = 0.454 kg; 1 lb/MMBtu = 0.430 g/MJ.) Emissions-to-load relationships were studied for each individual unit in Colorado, New Mexico, Utah and Wyoming. For power plants outside of those four states, we estimated these relationships individually for the largest power plants, accounting for 70% of the generation. For the remaining power plants, we applied average emissions factors by plant type – either coal, natural gas combined cycle, or natural gas combustion turbine. As illustrated for NOx emissions in Appendix C Figs. C1 and C2, for individual units the relationship between NOx or SO2 emissions and power plant gross load is generally linear (but with a non-zero intercept) when the load is higher than 40% of maximum gross load in the case of coal power plants and above 50% of the maximum gross load in the case of natural gas power plants. Especially for natural gas power plants, however, NOx emissions are relatively high per unit generation at low load, compared to stabilized emissions rates. To account for this effect, we estimated separate NOx emissions versus load relationships
For each scenario, a yearlong simulation was performed with daily steps of 24 h-long intervals, which included hourly regional level load data and unit level renewable generation data. Fig. 2 shows the total annual generation broken down by type for each scenario. As shown, the natural gas prices, greenhouse gas fees and availability of renewables clearly affect the generation mix. As expected, natural gas generation is highest in the cheap gas scenario, accounting for 22% of generation. Natural gas accounts for 12% of generation in the GHG fees scenario (which has significantly more wind generation compared to
Fig. 2. Annual generation by type for four different scenarios within the Western Interconnection.
604
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
Fig. 4. Time series generation dispatch stack for all of the Western Interconnection for the first week of July (typical summer week). Fig. 3. Annual generation by type for geographic sub-regions within the Western Interconnection.
the other scenarios, based on the MARKAL model outputs for wind capacity). Coal provides about 25% of generation in the base case, declining to 17% in the GHG fees case. Modest curtailment of renewables is seen in the GHG fees scenario, which has the highest wind capacity. Fig. 3 shows the breakdown of annual generation for sub-regions within the Western Interconnection area for the cheap gas, costly gas, and GHG fees cases. (The base case is omitted because results are similar to those for the cheap gas case.) Similar to what was shown in Fig. 2, in Fig. 3 the largest differences between the scenarios are the changes in wind generation, specifically in CA, CO + WY, and WA + OR + ID, and the reductions in natural gas and coal generation in AZ, CA, CO + WY and WA + OR + ID. Total generation increases in CO + WY in the GHG fees scenario, as wind power is exported to other areas. There is also increased curtailment in CO + WY and CA in the GHG fees case, which suggests that either the system cannot fully utilize the available generation, there is no available transmission to export it, or both. Committed fossil fuel generator minimum generation levels can limit the amount of local renewable generation that can be used since they cannot be decommitted. If nearby transmission lines are at their flow limits or the areas where the power needs to be exported are already congested, that renewable generation must be curtailed. Air quality implications of electricity generation are not just determined by the annual generation mix, but also by the generation mix over shorter time scales. To illustrate the short-term variations, Fig. 4 shows an hourly dispatch stack for the entire Western Interconnection for the first week in July for the cheap gas, costly gas, and GHG fees scenarios. (Again the base case is excluded because results are similar to those for the cheap gas case.) This is a typical summer time period, with relatively high load and lower wind resources than other times of year. During this time period, coal accounts for 22% of generation in the cheap gas case, 23% in the costly gas case, and 21% in the GHG fees case. The contribution from natural gas is 25% in the cheap gas case, 17% in the costly gas case, and 16% in the GHG fees case. Towards the end of the week the wind resource increases, decreasing the amount of
Fig. 5. Time series generation dispatch stack for the Colorado and Wyoming sub-region for the first week of July (typical summer week). Dashed line shows load in the subregion.
natural gas and coal used in the costly gas and GHG fees cases. In the GHG fees case, wind generation displaces coal through more frequent and larger magnitude ramping. Fig. 5 shows another level of detail for the hourly dispatch in July, focusing in on the CO + WY region (Public Service Company of Colorado (PSCO), PacifiCorp East – Wyoming (PAWY), and Western Area Power Administration, Colorado-Missouri Region (WACM) balancing
605
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
Table 2 Annual emissions from coal and natural gas fired generating units for the U.S. portion of the Western Interconnection. Percent reductions from the baseline case are shown in parentheses. Scenario
CO2 from Coal (million metric tons)
CO2 from NG (million metric tons)
Combined CO2 (million metric tons)
Baseline Cheap Gas Costly Gas GHG Fees
206 192 (7%) 173 (16%) 134 (35%)
55 68 (−24%) 32 (42%) 28 (49%)
261 260 (0.4%) 205 (21%) 162 (38%)
Scenario
SO2 from Coal (thousand metric tons)
SO2 from NG (thousand metric tons)
Combined SO2 (thousand metric tons)
Baseline Cheap Gas Costly Gas GHG Fees
114 104 (9%) 94 (18%) 71 (38%)
0.3 0.4(−33%) 0.2(33%) 0.2(33%)
114 104 (9%) 94 (18%) 71 (38%)
Scenario
NOx from Coal (thousand metric tons)
NOx from NG (thousand metric tons)
Combined NOx (thousand metric tons)
Baseline Cheap Gas Costly Gas GHG Fees
183 167 (9%) 149 (19%) 111 (39%)
12 16 (−33%) 7 (42%) 6 (50%)
196 183 (7%) 156 (20%) 118 (40%)
Fig. 6. Hourly NOx emissions projections from coal and natural gas power plants in the U.S. portion of the Western Interconnection for the cheap gas, costly gas and GHG fees scenarios.
18% lower in this case than in the base case. Total SO2 emissions in the GHG fees scenario are about 38% those in the base scenario. Emissions of NOx are also dominated by coal-fired generation, with emissions from natural gas-fired generation accounting for less than 10% of the total in all scenarios. Compared to the base case, total annual NOx emissions are reduced by about 7% in the cheap gas case, 20% in the costly gas case, and 40% in the GHG fees case. However, as shown below, the differences between the scenarios vary significantly over the year. Fig. 6 shows the area-wide hourly NOx emissions over the year for the cheap gas, costly gas, and GHG fee scenarios. (Again, results for the base case are similar to those for the cheap gas case.) For coal, while the GHG fee scenario clearly has lower emissions than other scenarios, the differences are most pronounced in spring and fall. During the summer season, when NOx emissions contribute most to ozone formation, the combined emissions from coal-and natural gas-fired power plants are similar across all four scenarios. In the July to September period, electricity demand increases and availability of wind decreases so that many coal and gas-fired power plants must be dispatched, regardless of scenario. As expected, NOx emissions from natural gas power plants are highest in the cheap gas scenario. These emissions show a pronounced summer peak, again corresponding to high load and relatively low wind resource availability. In particular, while generation from natural gas combined cycle plants occurs year-round in all the scenarios, operation of less efficient and higher-emitting combustion turbines is mainly limited to the summer peak period. As noted above, compared to the base case, total annual power plant NOx emissions across the region are reduced by about 7% in the cheap gas case, 20% in the costly gas case, and 40% in the GHG fees case. However, the differences across cases are sharply curtailed if only the summer months are considered. For example, averaged over July and August, power plant NOx emissions in the cheap gas, costly gas, and GHG fees cases are only 2%, 4%, and 8% lower than those in the base case. This dampened effect is shown in more detail in Fig. 7, which presents total NOx emissions from coal and natural gas-fired electricity generation for the first week of July. Results for the U.S. portion of the Western Interconnection are shown in Fig. 7a; results for CO + WY in Fig. 7b. During this time period, hourly emissions rates for the cheap and costly gas cases cross, depending on the day and time. This occurs for both the Western Interconnection domain and the CO + WY subregion. On several days during this period, NOx emissions from natural gas plants in the costly gas case are about equal to those in the GHG fees
areas). This area shows larger differences between scenarios than the plot for the entire interconnection. As expected, the cheap gas case has the most natural gas generation. In the costly gas case, wind generation is much larger and almost completely displaces natural gas except for the evening peak load hours and when wind generation decreases. Coal generation is similar to that in the cheap gas case, although there is an increase in ramping frequency. The GHG fees case has more wind generation and less coal generation than any of the other scenarios. Similar to the costly gas case, there is little natural gas generation. Coal generation ramping is increased in frequency, ramp rate, and overall magnitude and appears to be independent of load. 4.2. Emissions Table 2 shows the consequences of each of the scenarios for annual SO2, CO2 and NOx emissions from electricity generation in the Western Interconnection region, excluding portions in Mexico and Canada. Emissions are impacted by both long-term decisions (the type of generation that is built) and short-term decisions (how different generators are operated). In the cheap gas case, emissions reductions relative to the base case are primarily due to running gas generators more than coal. In the costly gas and GHG fees scenarios, emissions reductions are primarily due to increased penetration of renewables. As expected, the GHG fees scenario has the lowest CO2 and SO2 emissions. Total CO2 emissions are similar in the base and cheap gas cases, as increased emissions from natural gas generation offset reductions in emissions from coal. Total CO2 emissions are about 20% lower in the costly gas case than in the base and cheap gas cases; increased renewable penetrations lead to sharp reductions in natural gas use combined with more modest reductions in generation from coal. Total CO2 emissions in the GHG fees case are about 38% lower than those in the base and cheap gas cases due to renewable generation displacing both gas and coal. SO2 emissions in all cases are dominated by coal-fired generation, as other resources have low or no emissions. Correspondingly, SO2 emissions are about 9% lower in the cheap gas case than in the base case, as less coal is used in the cheap gas case. Because wind displaces some coal as well as natural gas in the costly gas case, SO2 emissions are about 606
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
corresponding air pollutant emissions in the western U.S., circa 2030. In contrast to typical emissions scenario analyses that consider annual and regionally averaged results, this study focuses on understanding how emissions responses might differ within the region and over the year. Although averaged results are adequate for CO2 emissions, higher spatial and temporal resolution is needed to understand local air quality implications for SO2 and NOx emissions. We focused here on NOx emissions due to the contribution this pollutant makes to elevated ozone concentrations in the western U.S. The results of our study show the potential for significant shifts in the electricity generation mix across the region, depending on fuel prices and whether or not emissions fees are applied. Fuel prices and fees specified for each scenario in the prior study with MARKAL (McLeod et al., 2014) determine the renewable energy capacity assumed for the region, which is input to the dispatch model. Across the region, electricity from wind ranges from less than 10% of generation in the base and cheap gas cases to 26% in the GHG fees case. Solar energy contributes about 7% of generation in all of the scenarios. With the renewables levels given, fuel prices and fees further influence the dispatch split between natural gas and coal. Natural gas contributes from 12% of annual generation in the region in the GHG fees case to 22% of generation in the cheap gas case. Coal-fired power plants contribute from 17% of generation in the GHG fees case to 25% in the base case. The generation mix and scenario responses show pronounced differences across the Western Interconnection region. Natural gas generation occurs mainly in CA and the WA + OR + ID sub-region, especially in the base (not shown) and cheap gas scenarios. Coal generation is used most heavily in AZ and CO + WY, in the base, cheap, and costly gas scenarios. Wind displaces coal in CO + WY, with wind generation increasing for export to other sub-regions in the costly gas and especially the GHG fees cases. Generation shifts between wind and natural gas on the West Coast. Responses in the four scenarios also vary sharply across the year. In particular, during summer, relatively full utilization of coal and natural gas plants is required to meet high demand, regardless of fuel prices and emissions fees. Annual results for the U.S. portion of the Western Interconnection show reductions in CO2 emissions of < 1%, 20% and 38% in the cheap gas, costly gas, and GHG fees cases, respectively, compared to the base case. Region-wide annual reductions in SO2 emissions are 9%, 18% and 38% in the cheap gas, costly gas, and GHG fees cases, respectively; reductions in NOx emissions are 7%, 20%, and 40%. Ignoring the difference in start-up versus stable NOx emissions rates for natural gas plants has little impact on region-wide emissions estimates, but would be more important in the local vicinity of some of these units. A key insight from downscaling the emissions estimates is that NOx emissions rates are highest in summer in all scenarios, due to relatively heavy reliance on coal and natural gas in that season. The seasonal enhancement is greatest for the GHG fees case, in which reductions compared to other cases occur in spring, fall, and winter, but not in the summer. In that case, across the U.S. portion of the Western Interconnect, the average NOx emissions rate in July and August is more than 60% higher than the average emissions rate for the year. Furthermore, during the summer period the NOx emissions reductions that are seen occur mainly at night. Accounting for seasonal and diurnal timing is important, because NOx emissions reductions are most effective at reducing ozone pollution if they occur on hot, stagnant summer days when meteorological conditions are most conducive to the formation and build-up of ozone from photochemical processes (NRC, 1991; Mesbah et al., 2015). Our study thus suggests that ozone pollution benefits might be overestimated if policy analyses rely on emissions reduction estimates with only annual time resolution. Emissions reductions that occur outside the summer season are still beneficial, however, because NOx emissions also contribute to regional haze, to health effects from secondary aerosols, and to nitrogen deposition that harms high altitude ecosystems in the western U.S. Results of this scenario analysis are not intended to be taken as
Fig. 7. Hourly total NOx emissions (coal and natural gas combined) in (a) the U.S. portion of the Western Interconnection and (b) the CO + WY sub-region for a typical summer week.
case and are less than half those in the cheap gas case (separate results not shown). However, because coal-fired power plant emissions dominate overall, the difference in natural gas plant emissions is offset by coal-plant emissions that are modestly lower in the cheap gas case. Again considering either the whole region or the CO + WY sub-region, NOx emissions reductions in the GHG fees case during this week occur mainly at night. In contrast, reductions in afternoon peak period emissions are comparatively modest. Results shown for NOx emissions in Table 2 and in Figs. 6 and 7 were computed using differentiated emissions versus load relationships for start-up conditions. The effect of doing so is shown in Appendix C, considering both annual and July/August averaging periods. Overall, when the calculations are simplified by using average relationships encompassing start-up periods together with stabilized loads, total NOx emissions are underestimated by less than 1% in all cases. However, NOx emissions from natural gas-fired power plants are underestimated by up to 14%, which occurs for the GHG fees case in July/August. 5. Conclusions and policy implications This study examines how future scenarios for natural gas prices and GHG emissions fees might influence the electricity generation mix and 607
Energy Policy 109 (2017) 601–608
R. Nsanzineza et al.
energy source, Energy Information Administration, 〈https://www.eia.gov/ electricity/data/state/〉 (Accessed 14 July 2016). EIA, 2016e. How much carbon dioxide is produced per kilowatthour when generating electricity with fossil fuels? Energy Information Administration. 〈www.eia.gov/ tools/faqs/faq.cfm?Id=74&t=11〉 (Accessed 16 January 2017). EIA, 2016f. Annual energy outlook 2016: with projections to 2040, Energy Information Administration, 〈https://www.eia.gov/outlooks/archive/aeo16/〉 (Accessed 9 February 2017). EPA, 2014. Regulatory impact analysis for the proposed carbon pollution guidelines for existing power plants and emission standards for modified and reconstructed power plant. U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards, Research Triangle Park, NC. EPA, 2016. Cross-state air pollution rule update for the 2008 ozone NAAQS, 81 Fed. Reg. 74504, October 26. EPA, 2016a. 2015 National Ambient Air Quality Standards (NAAQS) for Ozone. U.S. Environmental Protection Agency. 〈https://www.epa.gov/ozone-pollution/2015national-ambient-air-quality-standards-naaqs-ozone〉 (Accessed 14 July 2016). EPA, 2016b. Air markets program data, U.S. Environmental Protection Agency. 〈https:// ampd.epa.gov〉 (Accessed 16 January 2017). Eurek, K., Cole, W., Bielen, D., Blair, N., Cohen, S., Frew, B., Ho, J., Krishnan, V., Mai, T., Sigrin, B., Steinberg, D., 2016. Regional Energy Deployment System (ReEDS) Model Documentation: Version 2016, NREL/TP-6A20-67067. National Renewable Energy Laboratory, Golden, CO. Gilbraith, N., Powers, S.E., 2013. Residential demand response reduces air pollutant emissions on peak electricity demand days in New York City. Energy Policy 59, 459–469. http://dx.doi.org/10.1016/j.enpol.2013.03.056. Hobbs, B., Hu, M.-C., Chen, Y., Ellis, J.H., Paul, A., Burtraw, D., Palmer, K.L., 2010. From regions to stacks: spatial and temporal downscaling of power pollution scenarios. IEEE Trans. Power Syst. 25 (2), 1179–1189. IWG, 2013. Technical support document: technical update of the social cost of carbon for regulatory impact analysis, Interagency Working Group on Social Cost of Carbon, Washington, DC. Kerl, P.Y., Zhang, W., Moreno-Cruz, J.B., Nenes, A., Realff, M.J., Russell, A.G., Sokol, J., Thomas, V.M., 2015. New approach for optimal electricity planning and dispatching with hourly time-scale air quality and health considerations. Proc. Natl. Acad. Sci. USA 201413143. Krieger, E.M., Casey, J., Shonkoff, S.B.C., 2016. A framework for siting and dispatch of emerging energy resources to realize environmental and health benefits: case study on peaker power plant displacement. Energy Policy 96, 302–313. Lenox, C.D., Gage, C., Kaplan, O., Loughlin, D., Yelverton, W., 2013. EPA U.S. Nine-region MARKAL database: Database documentation. Environmental Protection Agency, Cincinnati, OH, U.S. Martinez-Anido, C.B., Botor, B., Florita, A.R., Draxl, C., Lu, S., Hamann, H., Hodge, B., 2016. The value of day-ahead solar power forecasting improvement. Sol. Energy 129, 192–203. McLeod, J., Brinkman, G., Milford, J.B., 2014. Emissions implications of future natural gas production and use in the U.S. and in the Rocky Mountain region. Environ. Sci. Technol. 48, 13036–13044. Mesbah, S.M., Hakami, A., Schott, S., 2015. Optimal ozone control with inclusion of spatiotemporal marginal damages and electricity demand. Environ. Sci. Technol. 2015 (49), 7870–7878. http://dx.doi.org/10.1021/acs.est.5b01178. Novacheck, J., Johnson, J., 2017. Diversifying wind power in real power systems. Renew. Energy 106, 177–185. NRC, 1991. Rethinking the Ozone Problem in Urban and Regional Air Pollution. National Research Council, Committee on Tropospheric Ozone. National Academy Press, Washington, DC. Pacsi, A., Alhajeri, N.S., Zavala-Araiza, D., Webster, M.D., Allen, D.T., 2013. Regional air quality impacts of increased natural gas production and use in Texas. Environ. Sci. Technol. 47 (7), 3521–3527. http://dx.doi.org/10.1021/es3044714. Pacsi, A., Kimura, Y., McGaughey, G., McDonald-Buller, E.C., Allen, D.T., 2015. Regional ozone impacts of increased natural gas use in the texas power sector and development in the Eagle Ford Shale. Environ. Sci. Technol. 49, 3966–3973. http://dx.doi.org/10. 1021/es5055012. Plachinski, S.D., Holloway, T., Meier, P.J., Nemet, G.F., Rrushaj, A., Oberman, J.T., Duran, P.L., Voight, C.L., 2014. Quantifying the emissions and air quality co-benefits of lower-carbon electricity production. Atmos. Environ. 94, 180–191. Thompson, T.M., King, C.W., Allen, D.T., Webber, M.E., 2011. Air quality impacts of plugin hybrid electric vehicles in Texas: evaluating three battery charging scenarios. Environ. Res. Lett. 6, 024004. Thompson, T.M., Rausch, S., Saari, R.K., Selin, N.E., 2014. A systems approach to evaluating the air quality co-benefits of US carbon policies. Nat. Clim. Change 4, 917–923. Thompson, T., Rausch, S., Saari, R.K., Selin, N.E., 2016. Air quality co-benefits of subnational carbon policies. J. Air Waste Manag. Assoc. http://dx.doi.org/10.1080/ 10962247.2016.1192071. Trail, M., Tsimpidi, A.P., Liu, P., Tsigaridis, K., Rudokas, J., Miller, P., Nenes, A., Hu, Y., Russell, A.G., 2014. Sensitivity of air quality to potential future climate change and emissions in the United States and major cities. Atmos. Environ. 94, 552–563. WECC, 2015. TEPPC study report – 2024 PC1 common case and release notes for WECC 2024 common case version 1.1, Western Electricity Coordinating Council, 〈https:// www.wecc.biz〉.
projections of future emissions. Rather, the study is meant to provide insight into potential trends and tradeoffs between contrasting scenarios, while recognizing limitations in their assumptions. These limitations include the use of a single historical year for meteorological inputs and the load profile, which provides an inherently limited representation of what future conditions might be like. We used the latest information on utilities’ plans that was available at the time this study was started, but those plans continue to evolve. The dispatch results are constrained by the electricity transmission infrastructure assumed in the model; future changes to this system would alter the results. Furthermore, fuel price forecasts are uncertain and evolving; more recent projections suggest future natural gas prices could fall below those assumed in our cheap gas scenario (EIA, 2016f). The air quality implications of the four electricity generation scenarios are being investigated in ongoing research that will use the detailed emissions results from this study in an atmospheric chemistry and transport model. Funding Support for this work was provided by the National Science Foundation (NSF) AirWaterGas Sustainability Research Network CBET1240584. Findings or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the NSF. Acknowledgments The authors thank Kent Kurashima for assistance with data analysis and graphics and Shannon Capps for feedback on early drafts of the results. Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.enpol.2017.07.051. References Bloom, A., Townsend, A., Palchak, D., Novacheck, J., King, J., Barrows, C., Ibanez, E., O’Connell, M., Jordan, G., Roberts, B., Draxl, C., Gruchalla, K., 2016. Eastern Renewable Generation Integration Study, NREL/TP-6A20-64472. National Renewable Energy Laboratory, Golden, CO. Brinkman, G.L., Denholm, P., Hannigan, M.P., Milford, J.B., 2010. Effects of plug-in hybrid electric vehicles on ozone concentrations in Colorado. Environ. Sci. Technol. 44, 6256–6262. Brinkman, G., Jorgenson, J., Ehlen, A., Caldwell, J., 2016. Low Carbon Grid Study: Analysis of a 50% Emission Reduction in California, NREL/TP-6A20-64884. National Renewable Energy Laboratory, Golden, CO. Brown, K.E., Henze, D.K., Milford, J.B., 2013. Accounting for climate and air quality damages in future US electricity generation scenarios. Environ. Sci. Technol. 47 (7), 3065–3072. Brown, K.E., Henze, D.K., Milford, J.B., 2017. How accounting for climate and health impacts of emissions could change the US energy system. Energy Policy 102, 396–405. Buonocore, J.J., Luckow, P., Norris, G., Spengler, J.D., Biewald, B., Fisher, J., Levy, J.I., 2015. Health and climate benefits of different energy-efficiency and renewable energy choices. Nat. Clim. Change 6, 100–107. http://dx.doi.org/10.1038/ NCLIMATE2771. Cole, W., Mai, T., Logan, J., Steinberg, D., McCall, J., Richards, J., Sigrin, B., Porro, G., 2016. 2016 Standard scenarios report: a U.S. electricity sector outlook, National Renewable Energy Laboratory Technical Report NREL/TP-6A20-66939, Golden, CO. EIA, 2013. Annual energy outlook 2013. Energy Information Administration, Washington, DC. EIA, 2016a. State energy data system, Energy Information Administration. 〈https://www. eia.gov/state/?Sid=US〉 (Accessed 14 July 2016). EIA, 2016b. Monthly energy review: coal, Energy Information Administration., 〈https:// www.eia.gov/totalenergy/data/monthly/〉 (Accessed 14 July 2016). EIA, 2016c. Natural gas data: natural gas withdrawals and production, Energy Information Administration, 〈https://www.eia.gov/naturalgas/data.cfm〉 (Accessed 14 July 2016). EIA, 2016d. Electricity, detailed state data, net generation by state by type of producer by
608