Policy Analysis pubs.acs.org/est
Emissions Reductions from Expanding State-Level Renewable Portfolio Standards Jeremiah X. Johnson*,† and Joshua Novacheck†,‡ †
Center for Sustainable Systems, School of Natural Resources & Environment, University of Michigan, 440 Church Street, Ann Arbor, Michigan 48109, United States ‡ Department of Mechanical Engineering, University of Michigan, Ann Arbor, Michigan 48109, United States S Supporting Information *
ABSTRACT: In the United States, state-level Renewable Portfolio Standards (RPS) have served as key drivers for the development of new renewable energy. This research presents a method to evaluate emissions reductions and costs attributable to new or expanded RPS programs by integrating a comprehensive economic dispatch model and a renewable project selection model. The latter model minimizes incremental RPS costs, accounting for renewable power purchase agreements (PPAs), displaced generation and capacity costs, and net changes to a state’s imports and exports. We test this method on potential expansions to Michigan’s RPS, evaluating target renewable penetrations of 10% (business as usual or BAU), 20%, 25%, and 40%, with varying times to completion. Relative to the BAU case, these expanded RPS policies reduce the CO2 intensity of generation by 13%, 18%, and 33% by 2035, respectively. SO2 emissions intensity decreased by 13%, 20%, and 34% for each of the three scenarios, while NOx reductions totaled 12%, 17%, and 31%, relative to the BAU case. For CO2 and NOx, absolute reductions in emissions intensity were not as large due to an increasing trend in emissions intensity in the BAU case driven by load growth. Over the study period (2015 to 2035), the absolute CO2 emissions intensity increased by 1% in the 20% RPS case and decreased by 6% and 22% for the 25% and 40% cases, respectively. Between 26% and 31% of the CO2, SO2, and NOx emissions reductions attributable to the expanded RPS occur in neighboring states, underscoring the challenges quantifying local emissions reductions from state-level energy policies with an interconnected grid. Without federal subsidies, the cost of CO2 mitigation using an RPS in Michigan is between $28 and $34/t CO2 when RPS targets are met. The optimal renewable build plan is sensitive to the capacity credit for solar but insensitive to the value for wind power.
1. INTRODUCTION Renewable Portfolio Standards (RPS) are state-level policies that have served as key drivers for the development of new renewable energy. While the policy designs can vary greatly across states, these programs typically mandate that a share of total electric sales must be met by qualified renewable resources, to be procured by local load serving entities. Currently, 29 states have binding RPS policies.1 While RPS policies are designed to increase renewable energy deployment, a common justification for their adoption is the abatement of pollutants and the potential for improved health outcomes. These claims are not always evaluated or quantified when states consider adopting or expanding RPS policies. Several methods of varying quality may be employed to evaluate the environmental benefits and program costs, but there is not a universally accepted method to evaluate the program’s impacts.2,3 Divergent results of such analyses, promoted by stakeholders when interests align, make it challenging to objectively evaluate the benefits and costs of these programs. The objective of this research is to present a rigorous method that comprehensively evaluates the future emissions reductions and costs attributable to new or expanded RPS programs. This approach employs a large-scale unit commitment and economic © 2015 American Chemical Society
dispatch model, coupled with a renewable project selection model that determines the least cost qualified renewable generation needed to meet the RPS targets. Taking into account renewable projects’ revenue requirements, offset energy costs, and capacity value, we select renewable projects to minimize the above-market costs, thereby minimizing RPS program costs. This approach appropriately values the time of day of generation, the capacity value of renewables, and the relevant market characteristics that may impact the RPS program. Several studies have evaluated the costs of RPS policies and their effectiveness at greenhouse gas mitigation. Heeter and colleagues3 conducted an extensive review of existing RPS cost and benefits estimates. Historical costs were based on the cost of renewable energy credits, alternative compliance payments, and direct reported from regulated utilities, with most reported cost impacts ranging between 0% and 4% of retail rates. Far fewer estimates of RPS benefits exist, with considerable diversity of methods and results. In a separate review of RPS Received: Revised: Accepted: Published: 5318
December 16, 2014 March 31, 2015 April 3, 2015 April 17, 2015 DOI: 10.1021/es506123e Environ. Sci. Technol. 2015, 49, 5318−5325
Policy Analysis
Environmental Science & Technology impacts, Chen and colleagues2 demonstrated that wind has been the dominant technology to meet RPS requirements and that the rate impacts to customers tend to be less than 1%. The authors call for several improvements in methods that are addressed in our study, including improved treatment of the value of capacity, more extensive use of scenario analysis, and the examination of the introduction of carbon prices. Other studies of RPS performance include an empirical investigation of the relationship between RPS implementation and share of renewables,4 analysis on the impacts of diversity of policy design,5 use of the long-term price elasticity of supply of renewables to evaluate CO2 abatement from RPS,6 and the deployment of a MARKAL framework to assess implications of high target RPS goals.7 Multiple studies have assessed the interaction between or comparison of energy policies, as they relate to RPS.8−11 The research presented here differs from these studies by coupling a robust unit commitment and economic dispatch model with an optimization model to minimize future RPS compliance costs. To demonstrate the utility of this method, we evaluate the environmental and cost impacts of three alternative expansions to Michigan’s RPS.
RRE , t = (1 + i)t × CPPA , t × Pt
where rWACC equals the weighted average cost of capital, RRE,t is the revenue requirement for the renewable project in year t in $, CRE,t is the cost of the renewable project in year t in $ inclusive of taxes, i is the assumed value for inflation, and Pt is the expected generation (MWh). The analysis is conducted in five-year increments from 2015 through 2035. The calculation for annual project costs, CRE,t, follows the approach detailed in Johnson et al.,13 which includes the costs associated with construction, investment costs, operations and maintenance, taxes, and insurance, discounted over the project life to provide a levelized value. The method can also include the impacts of federal subsidies, such as the Production Tax Credit (PTC) and Investment Tax Credit (ITC). This study assumes those benefits are not renewed and thus not available, but that assumption is tested in the sensitivity analysis. The costs associated with the renewable PPAs are offset by two factors in the site selection. The first factor is the offset energy costs, which are assumed to equal the variable costs that are offset by the introduction of the renewable generation, reflecting the value of the time of day of generation, as shown in eq 5
2. METHODS We use two models in an iterative process to determine power system behavior and the optimal selection of renewable projects. The unit commitment and economic dispatch model determines generator operations and market energy prices, reflecting key generator and system constraints. The renewable project selection model determines the least cost renewable options considering factors such as the time of day of resource availability, energy and capacity values, and project costs. The power system representation used in the dispatch model is updated to reflect the new renewable projects in five-year iterations, spanning 2015 through 2035. These models are discussed in more detail in sections 2.1 and 2.2, while section 2.3 details the assumptions used for the case study of Michigan. 2.1. Renewable Project Selection. The objective of the renewable project selection model is to determine the new renewable generation that will minimize RPS program costs, as dictated by eqs 1 and 2 n min β = Σi = 1[(CPPA , i − Ce , i − Ccap , i)*Pi ]
n
Ce , i =
∑h = 1 (Pi , h × Eh) n
∑h = 1 (Pi , h)
(5)
where Pi,h is the generation (MWh) from renewable project i in hour h, and Eh is the market energy price in hour h ($/MWh) in real terms, which is an output of the dispatch model discussed in section 2.2. Equation 6 determines the capacity costs offset by the renewable project kW
Ccap , i =
CNE*γi*103 MW CFi*8760 h/yr
(6)
where CNE is the cost of new entry ($/kW-yr) which represents the lowest cost new firm capacity that is available, γ represents the capacity credit (%) awarded to each technology type which represents its contribution to firm capacity as a function of its installed capacity, and CFi is the capacity factor of renewable project i. In this analysis, CNE equals the costs of a new combustion turbine with a 20-year lifetime, used solely for capacity purposes. In the Supporting Information, we provide an illustrative example of a renewable resource supply stack created using this method based on both revenue requirements and above-market costs. 2.2. Unit Commitment and Economic Dispatch Model. To determine generator operations and market energy prices, we use a comprehensive unit commitment and economic dispatch model. This analysis is conducted using Plexos for Power Systems by Energy Exemplar, running Xpress-MP 25.01.05 solver. Plexos employs linear programming with an interior point algorithm to determine the least cost dispatch for all generators across the system. To determine the on−off unit commitment decisions, we use mixed integer programming with branching and cutting methods, assuming commitment decisions are made with a 24-h look-ahead. After the units are committed, their dispatch is determined to minimize the system costs while considering operational constraints on generators
(1)
such that Σin= 1Pi ≥ φ
(4)
(2)
where β is the above-market costs ($) associated with the introduction of renewable energy, CPPA,i is the cost of a power purchase agreement (PPA) for renewable project i ($/MWh), Ce,i are energy costs offset by the introduction of renewable project i ($/MWh), Ccap,i are capacity costs offset by the introduction of renewable project i ($/MWh), Pi is the annual generation from qualified renewable project i (MWh), and φ is the total incremental renewable demand driven by the RPS targets and influenced by load growth (MWh). Qualified renewable projects must meet the technical and deliverability requirements for the state’s RPS. CPPA is determined such that discounted project costs equal discounted project revenues, as shown in eqs 3 and 412,13 ΣTt = 0(1 + rWACC)−t × RRE , t = ΣTt = 0(1 + rWACC)−t × CRE , t (3) 5319
DOI: 10.1021/es506123e Environ. Sci. Technol. 2015, 49, 5318−5325
Policy Analysis
Environmental Science & Technology
2.2.3. Fuels and Emissions. Baseline fuel price forecasts are based on the EIA’s 2014 Annual Energy Outlook for coal and natural gas.22 Zonal fuel price differences are determined using fuel costs reported in the EIA 923 database. Coal prices vary annually, while natural gas prices vary monthly. Uncontrolled emissions rates assumed for each fuel type are based on EIA and EPA assumptions.23,24 NOx emissions rates, which vary considerably among generators that use the same fuel, are generator-specific.25 SO2 removal efficiency is dictated by control technology type. 2.2.4. Load. We use contemporary load data and demand forecasts from Independent System Operators (ISOs) in regions where available18,19,26 and the Eastern Interconnection Planning Collaborative,14 dividing or aggregating load as needed to conform to zonal boundaries. Because intrazonal transmission is not modeled, the associated transmission losses are added to load, based on historical transmission loss rates from 2010 EIA State Profiles. 2.2.5. System Reliability. We met requirements for firm capacity and spinning reserves to ensure system reliability. Both up and down spinning reserves are added to contingency reserves at a rate of 3% of wind capacity to cover short-term wind variability, following a similar approach employed by National Renewable Energy Laboratory’s (NREL) Eastern Wind Transmission and Integration Study.27 To ensure that sufficient firm capacity is available to meet peak demand, we maintain a reserve margin of 14.2% across the constrained geography.28 Available capacity is the summer capacity of firm generation adjusted by the forced outage rate, plus any capacity credit awarded to variable renewables. We add natural gas combustion turbines to maintain capacity levels to meet the peak demand plus the reserve margin. Variable renewables receive a capacity credit (γ) of 14.1% of installed capacity for wind29 and 38% for solar.30 While the capacity credit for wind is based on the determination of firm capacity value from a MISO study, comparable values were not yet available from MISO for solar, so values from the neighboring PJM ISO are used. We test the sensitivity of the renewable build plan and the relative competitiveness of onshore wind versus utility-scale solar against the assumptions for variable renewable capacity credit. 2.3. Michigan Case Study. We test this approach on an expanded RPS for Michigan. Michigan’s current RPS requires 10% of load to be met by qualified renewables by 2015. This is expected to be met,31 after which no additional renewables are mandated. In addition to a business as usual (BAU) case, we examine three scenarios: 20% by 2030, 25% by 2025, and 40% by 2035. Linear interpolation is assumed between 2015 and the target completion date for each scenario, after which the share of renewables is maintained. (See the Supporting Information for a graphical representation of these targets.) We considered wind, utility-scale solar, distributed solar, municipal solid waste, landfill gas, and biomass from six feedstocks: crop residues, switchgrass, forest residues, primary mill, and urban wood. Wind resource data is based on 57 sites in Michigan.32 Solar generation uses typical meteorological year data and NREL’s System Advisor Model with six sites selected from a larger data set to ensure representation from each of Michigan’s Combined Statistical Areas, as well as one site representing the Upper Peninsula. We use biomass availability assumptions33 and estimate landfill gas resource based on EPA candidate landfills.34 We assume biopower projects operate at 80% capacity factor when calculating project viability.
and transmission, meeting energy and ancillary service demand in real time. The assumptions for generators, transmission, demand, ancillary services, fuel, and emissions are inputs into this model. The system is solved chronologically in hourly increments with a high degree of data resolution to fully capture the operational behavior of the system and the associated emissions. Details of key input assumptions are provided in the following sections, with additional information provided in the Supporting Information. 2.2.1. Zonal Representation. To effectively capture the import and export characteristics to and from Michigan we first modeled the full Eastern Interconnection. Then, we created a constrained geography that includes Michigan and neighboring zones, as shown in Figure 1. Imports into and exports out of the
Figure 1. Zonal boundaries for the full system representation, in grayscale, and the constrained system representation, in color.
constrained geography are held constant based on the full Eastern Interconnection modeling results. Within each zone, generators, fuel price, and load are modeled using the assumptions detailed in the following sections. Interzonal transmission interface limits are derived from Eastern Interconnection Planning Collaborative study and ISO studies.14−17 We treat transmission into Canada as fixed generation based on historical trade flows, with data from the ISOs.17−19 2.2.2. Generators. We use the EIA-860 database for characteristics of 10,000 generators, aggregating those less than 10 MW. Key assumptions for the generators include the heat rate (US EPA eGrid and Air Market Program data sets), variable operations and maintenance costs,20 and the presence of environmental control equipment (e.g., flue gas desulfurization). For each technology type, partial load heat rate impacts are included, and minimum operating capacities are assumed. Forced outage rates and mean time to repair are used to determine generator availability.21 For future year analysis, the portfolio of available generators changes through three drivers: coal unit retirements, new natural gas combustion turbines needed to meet capacity reserve requirements, and new renewable generators, as determined by the renewable project selection optimization. 5320
DOI: 10.1021/es506123e Environ. Sci. Technol. 2015, 49, 5318−5325
Policy Analysis
Environmental Science & Technology
Figure 2. Michigan generation mix by RPS scenario (a) business as usual, (b) 20% by 2030 case, (c) 25% by 2025 case, and (d) 40% by 2035 case, where BAU = business as usual, MSW = municipal solid waste, LFG = landfill gas, and PV = photovoltaic.
extension of federal tax subsidies (i.e., the PTC and ITC). Final sensitivities include two levels of CO2 tax; in “CO2 Tax 1,” the cost of CO2 is $25/t, held constant in real terms, while in “CO2 Tax 2,” a higher social cost of carbon with a 3% discount rate is used.41 The energy revenues that each project would receive are a function of the output of the dispatch model and the time of day of generation. We assume a value of capacity (CNE) of $90/ kW-yr, which is representative of the cost of new entry for a combustion turbine.42 Coal plant retirements include 1.2 GW of announced retirements and 750 MW of units that are not currently in compliance with the Mercury and Air Toxics Standards and cannot add control equipment at a cost less than a new natural gas combined cycle (as described in the Supporting Information). 2.3.1. Model Validation. We undertook several efforts to demonstrate the validity of both models. For the unit commitment and economic dispatch model, a simulation representing the power system in 2013 was conducted. We then compared the results for generation by fuel type and the seasonal diurnal variation of locational marginal price (LMPs) against actual values. As shown in the Supporting Information, the model closely captures the contribution from each of the main types of generators, as well as reflects seasonality and time of day price impacts. Accurately capturing power price seasonality and differences in off-peak and on-peak prices is essential for appropriately valuing variable resources such as solar and wind. To ensure the validity of the renewable site selection optimization results, we compared the value of the least cost PPA against recent actual PPAs signed in Michigan. We find wind projects available at $56/MWh (with PTC), while PPAs for wind power signed by DTE Energy in 2012 and 2013 ranged from $49/MWh to $53/MWh.31 This comparison
In 2014, we assume the following installed costs: $1,940/kW for onshore wind, $2,453/kWac for utility-scale solar photovoltaics, $4,734/kW-AC for distributed-scale solar photovoltaics, $4,505/kW for biomass and municipal solid waste, and $1,816/kW for landfill gas projects, based on several sources and converted to current dollars.35−38 We hold installed costs for wind, biomass, and landfill gas projects constant in real dollars, while the installed costs of solar decreases at rates equal to recent historical decreases (15% per year for utility-scale and 9% per year for distributed-scale PV37) until targets of $1,000/kWac (in 2013$) for utility-scale and $1,500/kWac (in 2013$) for distributed-scale solar are achieved. (See the Supporting Information.) Michigan fuel costs, in 2013$, for natural gas increase from $4.54/MMBtu in 2015 to $7.51/MMBtu in 2035, while assumed coal prices increase from $2.68/MMBtu in 2015 to $3.22/MMBtu in 2035, with base values from EIA 923 database and growth rates based on EIA’s Annual Energy Outlook.22 The forecasted load growth in Michigan is 0.485% per year, based on a Public Service Commission planning study,39 while the average rate of increase in the annual peak is 0.436%, based on MISO’s forecasts for Local Resource Zone 7.40 For each year, the deviation from the historical hourly shape is minimized, constrained by the annual load and peak demand. We conducted several sensitivities on the installed costs of renewables. In the “Low Solar Costs” sensitivity, utility-scale solar installed cost declines at the same rate as the base case but continues to decrease until reaching a floor of $500/kWac. We designed this sensitivity to represent an extreme example in which very low solar installed costs are achieved. In the “Low RE Cost” sensitivity, wind installed costs are 25% lower than base case assumptions, in addition to the low cost solar. In “High RE Cost,” wind’s installed cost is $2,425/kW, while utility-scale solar’s price floor is $1,250/kWac. We also conducted a sensitivity that examines the impact of an 5321
DOI: 10.1021/es506123e Environ. Sci. Technol. 2015, 49, 5318−5325
Policy Analysis
Environmental Science & Technology
Figure 3. (a) CO2, (b) SO2, and (c) NOx emissions intensities for four scenarios: business as usual (“BAU”), 20% by 2030, 25% by 2025, and 40% by 2035.
Figure 4. In-state and regional cumulative (a) CO2, (b) SO2, and (c) NOx reductions due to Michigan RPS, 2015−2035.
displace in-state gas generation (0.6 TWh, 0.8 TWh, 1.9 TWh), and increase net exports (4.0 TWh, 5.6 TWh, 13.4 TWh). Figure 3 shows the changes in emissions intensity of generation within Michigan for CO2, SO2, and NOx. The differences in net interstate imports between the scenarios make emissions intensity of in-state generation a more useful metric than total emissions within the state. The existing RPS for Michigan is assumed to achieve 10% renewable penetration (inclusive of existing incentives) by 2015, which is reflected in all cases. Therefore, the 20% by 2030% case adds renewables equal to 10% of retail sales; the 25% by 2025 case adds renewables equal to 15% of retail sales; and the 40% by 2035 adds renewables equal to 30% of retail sales. As shown in Figure 3a, by 2035, these three cases reduce the CO2 intensity of generation by 13%, 18%, and 33%, respectively. Due to load growth, over the study period (2015 to 2035), the absolute CO2 emissions intensity increased by 1% in the 20% RPS case and decreased by 6% and 22% for the 25% and 40% cases, respectively. The increase in emissions intensity beginning in 2025 for the 25% scenario and in 2030 for the 20% scenario represents the years in which each of those scenarios meet their renewable targets and begin simply maintaining their share of renewables. Figure 3b shows a precipitous drop in SO2 emissions across all cases due to compliance with the Mercury and Air Toxics Standards for which the control technologies have the cobenefit of SO2 reductions, diminishing the reduction potential that can be achieved through increased RPS targets. NOx emissions (Figure 3c) follow a reduction trend similar to CO2.
demonstrates that the renewable project selection model presents a realistic view on the resource costs in Michigan.
3. RESULTS Figure 2 shows the annual generation mix for each of the cases. The BAU case (Figure 2a) is dominated by coal and nuclear, with modest contributions by natural gas and wind. By doubling the RPS target from 10% to 20% by 2030, as shown in Figure 2b, increased onshore wind generation displaces coal and some natural gas. Increasing the RPS target to 25% and accelerating the timeline to reach that goal (Figure 2c) increases onshore wind generation. Figure 2d shows a 40% RPS, where utility-scale solar is selected over additional onshore wind in the final years of the study. Driven by installed costs and resource quality, onshore wind is the dominant technology for new renewable generation. In the 20% and 25% RPS scenarios, onshore wind makes up more than 97% of new renewable generation in 2035. Only in the 40% by 2035 case does utility-scale solar effectively compete with onshore wind, contributing 29% of new renewable generation. Solar’s competitiveness in this case is due in part to lower costs in the later years of the study, as well as depressed off-peak energy prices due to high penetrations of wind. New renewable resources also offset out-of-state generation through increased exports. In 2035, new renewable generation totaled 12.1 TWh, 17.9 TWh, and 34.9 TWh for each of the three scenarios. This renewable generation served to displace in-state coal generation (7.6 TWh, 11.5 TWh, 19.6 TWh), 5322
DOI: 10.1021/es506123e Environ. Sci. Technol. 2015, 49, 5318−5325
Policy Analysis
Environmental Science & Technology
load heat rates and different generators operating, which drive a small (
