2.3 Architecture
2.3.1 Existing Fleet
Aggregation
Although the existing generation fleet includes thousands of units, computational tractability within the multi-region inter-temporal optimization framework requires aggregation. For each region, fossil units with similar attributes were aggregated into blocks based on the principle that all units in a given block would be dispatched similarly based on market conditions. Since the most important factor for determining dispatch order is energy cost, units within each fuel group were classified according to heat rate. Additionally, gas turbine units and single cycle steam units were treated separately from combined cycle units. The aggregation process was designed to be flexible so that different schemes can be applied depending on the objectives of the analysis. For example, for an analysis of the value of retrofitting existing coal units with environmental controls, the coal units were aggregated by region in terms of the cost of retrofitting, for a total of twelve blocks per region. The current default formulation uses four coal blocks, seven gas blocks, and one petroleum block per region to represent the existing fossil fleet.
For each aggregated capacity block, the values of relevant performance attributes, such as heat rate and variable operating costs, were determined by taking the capacity weighted average of the parameter of interest across units assigned to that capacity block. In addition, an age distribution was calculated for each block of existing capacity based on the online year of the constituent units. Combined with an assumption about total lifetime after which a unit must retire, this translates into a "survival fraction" across model time periods of the block's initial capacity endowment. This distribution effectively creates a set of vintages within each existing capacity block for the purposes of retirement schedule, although the cost and performance characteristics (and therefore dispatch) are the same across vintages.
Existing nuclear capacity is represented by a single block per region with a similar age distribution approach. However, rather than a single lifetime, a percentage of capacity in each region (amounting to roughly 75% nationwide) was assumed to receive 20-year license extensions, taking their total lifetimes to 80 years, while the remainder were limited to 60-year licenses.[1] Announced retirements are incorporated explicitly where this occurs earlier than in the above calculation, see Table A-6 in the Appendices for details. The model also assumes a limited amount of uprate capacity (roughly 2.1 GW nationally) that slightly increases the existing nuclear endowment in the first decade of the model's time horizon.
Existing renewable capacity is also represented by a single block per technology per region, with blocks representing onshore wind, utility PV, rooftop PV, concentrated solar, hydro, geothermal, and biomass.[2] Wind, solar, and geothermal units have limited lifetimes, typically around 30 years, with the option to repower the site with a new unit. Biomass units have a 60 year lifetime. The hydro block has unlimited lifetime. Because most hydro capacity takes advantage of finite special resources, such as a geographically advantageous dam location, it may reasonably be expected neither to expand nor phase out over time.
Table 2‑1 describes the classification of existing generation.
| Generation | Number of Classes | Lifetime (years) | 2015 Total Capacity (GW) |
|---|---|---|---|
| Coal | 4 (based on heat rate) | 70 | 266 |
| Natural Gas (combined cycle) | 3 (based on heat rate) | 40 | 237 |
| Natural Gas (gas turbine) | 3 (based on heat rate) | 60 | 128 |
| Natural Gas (single cycle steam) | 1 | 70 | 75 |
| Petroleum | 1 | 70 | 37 |
| Nuclear | 1 | 60-80 | 98 |
| Hydro (incl. pumped storage facilities) | 1 | Indefinite | 79 |
| Wind | 1 | 25 | 76 |
| Geothermal | 1 | 30 | 3 |
| Solar (PV and Concentrated) | 1 | 30 | 31 |
| Biomass and Other | 1 | 60 | 18 |
A dataset on the existing electricity generating fleet was procured from Energy Velocity LLC based on data reported on Form EIA-860 (Annual Electric Generator Report) as of December 2015. This dataset characterizes the existing fleet at the unit level in terms of installed capacity, average heat rate, fuel costs, variable and fixed O&M costs, and unit age (online year). Additional data on emissions rates for SOx and NOx and announced retirement dates are procured from Energy Velocity LLC as of early 2019. These characteristics, with the exception of fuel costs, are held constant across the model time horizon. To enable a single national price for the various fuels that varies over time and scenario while also preserving the observed heterogeneity in realized costs, differences in fuel costs across units are represented relative to the (time-dependent) national average, and these differences themselves are assumed to diminish over time.
Retrofit Options for Existing Coal
The electric sector model includes several retrofit options for existing coal capacity in response to environmental policy. One retrofit option allows the unit (that is, all units within a capacity block) to improve its heat rate. The cost and improvement resulting from this is specific to a given analysis, and this feature is turned off for the base model. The model can also convert existing coal capacity blocks to biofuel or to natural gas, it can retrofit a unit to add carbon capture and storage, or it may add the capability to co-fire biomass at 5%. These transformations incur a one-time capital cost and can result in capacity degradation and changes in operating parameters including variable costs, heat rates, and emission rates (described below). Operating characteristics are expressed as relative to the associated base (i.e. pre-retrofit) coal plant. In addition, most retrofits result in a capacity penalty, which is at least as large as the heat rate penalty (since such retrofits do not expand the thermal capacity of the plant) and may be larger. Retrofits and conversions do not change the age profile of the underlying capacity.
Figure 2‑2 shows the possible retrofit endpoints and pathways. In its current configuration, the model starts with four existing coal capacity classes, all of which are assumed to be fully compliant with all existing non-CO2 regulations, but differ by heat rate. The four coal classes can undertake a carbon capture and storage retrofit to control CO2, retrofit to improve heat rate, convert to biofuel, convert to gas, or do nothing. They can also choose to retire. The existing units that retrofit to improve heat rate have the option to undertake carbon capture and storage retrofits in subsequent time periods. Cost and performance assumptions for these retrofit options are listed in Appendix A. At the time of writing, the heat rate improvement retrofit is inactive for lack of unit level data on the cost and potential for this option.
Model users may wish to analyze scenarios in which there exist physical limitations on the number of retrofits within a given time interval. Users have the ability to impose physical limits on retrofits for a given year(s), although no such limits are included in the baseline scenario. Such limits can be placed only at the national level, not the regional level.
2.3.2 New Generation and Storage Capacity
The model considers the following technologies when adding new generation and storage capacity.[3]
| Generation | Lifetime (years) |
|---|---|
| Supercritical Pulverized Coal (SCPC) without CCS (with full environmental controls) | 70 |
| Integrated Gasification Combined Cycle Coal (IGCC) without CCS | 60 |
| Integrated Gasification Combined Cycle Coal (IGCC) with 90% CCS | 60 |
| Integrated Gasification Combined Cycle Coal (IGCC) with 55% CCS | 60 |
| Natural Gas Combined Cycle (NGCC) without CCS | 60 |
| Natural Gas Combined Cycle (NGCC) with CCS (post-combustion capture) | 60 |
| Natural Gas Combustion Turbine | 60 |
| Dedicated Biomass | 60 |
| Dedicated Biomass with CCS | 60 |
| Nuclear (Generation III+) | 80 |
| Geothermal | 30 |
| Onshore Wind Turbines | 25 |
| Offshore Wind Turbines | 25 |
| Solar Photovoltaic (Fixed Tilt, Single Axis, and Double Axis) | 30 |
| Concentrating Solar Power (CSP) (solar thermal) | 60 |
| Lithium-Ion Battery | 20 |
| Compressed Air Energy Storage | 30 |
For new fossil generation, there is only one class (as opposed to the multiple classes used to describe the existing fleet). However, because cost and performance parameters are assumed to vary over time (described in the next section), separate vintages are maintained for new additions in each time period. For nuclear, there are two classes that vary only by capital cost. The first class represents new construction on "brownfield" sites (either a new unit co-located with existing units or a new plant at an existing site) and is assumed to be 10% less expensive than the second "greenfield" class. Capacity additions in the first class are constrained by region based on estimated availability of eligible brownfield sites. For wind and solar technologies, there are multiple classes by region simulating regional resource endowments. These are described in the next section. Storage technologies have one class per region, and are likewise described in the next section.
The core model uses cost and performance data for new electric generation and storage technologies based on published, publicly available EPRI reports (Program on Technology Innovation: Integrated Generation Technology Options, 2017). These reports focus on projections for the current and 10-15 year time horizon.[4] Details of these cost data are presented in Appendix A. There are minor differences in technology capital costs across regions, stemming primarily from differences in labor costs. Note that the capital costs do not include financing during construction, as these vary considerably depending upon the entity building the facility.
All new generation technologies are assumed to include all the required controls to comply with the new environmental regulations for air and water.
2.3.3 Carbon Capture and Storage
US-REGEN explicitly models regional carbon capture technology deployment and storage capacity. Retrofit and newly built capacity can include CCS technology that captures either 55 (for coal-fired units) or 90 percent (for natural gas-, coal-, and biomass-fired units) of emitted carbon. The captured CO2 is transported to injection sites where it is stored in saline aquifers. Regions must invest in CO2 injection capacity to enable CO2 storage, but regional storage capacity is limited based on estimates from the National Carbon Sequestration Database (NATCARB).[5] Once regional CO2 storage limits are reached, investments in inter-regional CO2 pipeline capacity can be made to access capacity in neighboring regions. A region's total CO2 storage requirement is described as the sum of all CO2 captured within the region net of the imports and exports. Since economies of scale are difficult to measure in a linear model and the cost of transport declines with increasing flow, pipeline costs were chosen conservatively. Inter-regional pipeline costs are equivalent to the costs of 20 percent utilization of a 42-inch pipeline, with each regional interconnection having a representative distance across which that pipeline must be constructed. Pipeline distances are calculated from a geographic centroid that is also used to calculate electricity transmission. Injection costs are calculated on a per state basis and each state has a fixed available capacity it cannot exceed on an annual basis. For all technologies, costs are a sum of capital investment and fixed operating and maintenance costs. The capital cost of CCS technologies includes the cost of a 20-mile CO2 pipeline that enables access to either a dedicated injection site or large pipeline for inter-regional transport.
Because regions' storage limits are effectively capped by the sum of their own and their eligible trading partners' in-region storage potential, some regions are limited in storing CO2 without pipeline costs. There is zero CO2 injection potential in fourteen states, but there are options to build pipelines to connect those states to available CO2 injection sites in neighboring states. These state CO2 storage capacities are listed in Table A-4 in Appendix A.
2.3.4 Availability and Variability
Availability factors were assigned to existing capacity through a de-rating process to capture average outages. In practice, units have their full capacity available in most hours, but are unavailable at certain times due to scheduled and unscheduled maintenance. Planned downtime is typically scheduled seasonally to coincide with lower expected loads. Thus a flat de-rating of capacity will underestimate availability at peak times, when unit operators plan for full capacity and are affected only by stochastic failure events. On the other hand, ignoring scheduled outages will overestimate total availability and therefore generation, particularly in the case of coal. Ideally, an hourly or daily availability factor could be inferred from reported unit operation data, but there are no data at this level of resolution. The closest approximation in publicly available EIA data is a table of monthly generation totals by state, and hence region, for aggregated fuel groups (e.g. coal, gas, nuclear, etc.) in historic years.
For existing nuclear and biomass / other capacity, a monthly availability factor was calculated to satisfy the equation
capacity × availability × hours in the month = generation in the month
for the calibration year 2015. This simple calculation was based on the idea that these units are "in the money" (i.e. have infra-marginal dispatch cost) all the time and hence not subject to dispatch decisions. For nuclear, with its very low variable costs, this is usually literally true; for the small amounts of existing biomass capacity, it is likely not literally true but it is a convenient simplifying assumption that will not strongly affect results. The average availability of existing nuclear units in 2015 was 93%. After 2015, both existing and new nuclear units are assumed to have a similar monthly shape and availability factor.
For fossil units, this calculation must be modified to include only "hours in the money"—the availability factor is intended to represent the average percentage of the time full capacity is available, which is likely greater than its dispatched capacity factor for high-variable cost units.[6] Because the choice of availability factor affects dispatch decisions, which affects "hours in the money" for the various units, these factors cannot be directly observed. Based on inspection of monthly generation patterns and a series of calibration experiments, we apply exogenous availability factors for coal and combined cycle natural gas that peak in the summer and winter and have annual averages of around 75% and 85% respectively. Gas turbines are assumed to have 100% availability.
For geothermal, a similar approach is taken as for nuclear (matching capacity to monthly generation using all monthly hours). The availability factor of geothermal is assumed to increase over time up to a maximum of 80%.
Hydro and pumped storage must match monthly generation constraints but are allowed to dispatch within the month subject to a minimum generation constraint that approximates stream flow and reservoir constraints. Because base year monthly generation of hydro may reflect anomalous water availability conditions, we use the long-run average monthly pattern to constrain generation in the model projection periods. Hourly dispatch of hydro resources is estimated in between model iterations with residual load serving as a proxy for the system price. This ensures that hydro dispatch coincides with peak residual load to the extent possible within the resource constraints. The resulting hourly profile of hydro dispatch is translated to the dynamic model via the weighted representative hours.
Availability factors are assigned to new thermal units based on the best-performing existing units.
For wind and solar power generators (existing and new), variability coefficients are based on hourly resource data as discussed below in Section 2.4.1.
2.3.5 Load
Load growth and hourly load shape inputs vary depending on whether the model is run in integrated or electric-only mode. In an integrated model run, regional hourly load shapes are calculated by the end-use model for each time step, as described in Section 3.6. In this case, the load shapes are updated for each model year to reflect changes in the patterns of energy demand over time.
In the electric-only model, the regional shapes are based on historical hourly load profiles, which are derived from a dataset from ABB Energy Velocity which assembles data from EIA reported by balancing authority. These shapes are translated to hourly shapes for each region and scaled to match total base year electricity consumption (retail and direct use) reported by the EIA. In this case, the load shape stays constant over the duration of the model. The reference load growth path for future time periods is based on the EIA's Annual Energy Outlook (AEO) but the model can be run under a variety of alternative load growth scenarios.
2.3.6 Inter-Regional Transmission
The electric model allows for inter-regional transfers of power using a bubble-and-spoke representation. In each segment or dispatch period, the amount of power transfer between a pair of regions is limited by the total transmission capacity between them.[7] The source for base year data is currently the inter-regional transmission capacity used by the National Renewable Energy Laboratory's ReEDS model (NREL 2017), except for lines starting or ending in New York, which are informed by NYISO's 2016 Reliability Needs Assessment (NYISO, 2016). An alternative set of assumptions derived from EPA's IPM 5.13 model (EPA 2017) is occasionally used to test the sensitivity of model results to transmission assumptions. These data are mapped to the model regions used by US-REGEN. Figure 2‑4 shows the existing power transfer capacity between the default 16 regions. Power flows between regions are subject to line losses based on the distance between the two regions. For every 100 miles of long-distance transmission, US-REGEN assumes 0.5% power losses. US-REGEN allows additional transmission capacity to be installed between regions at a cost of $3.85 million per mile for a notional high voltage line (e.g. 500kV AC or 800kV DC) to transfer 6400 MW of capacity.
The electricity transmission system is complex in terms of planning, operation, and organization. In many scenarios, including the reference scenario, additional build limits are imposed to reflect difficulties with siting or public acceptance. These limits have been generated in conjunction with experts from EPRI's Power Delivery and Transmission sector and take two forms. The first is a national constraint on total new transmission builds in GW-miles, equal to a 10% increase in the first decade, a 15% increase in the second decade, and a 20% increase in the third decade and beyond, relative to 2010 transmission capacity. There is an optimistic version of this constraint which doubles these percentages. These numbers are drawn from considering historical build rates for new transmission capacity. The second constraint is a limit on the amount of new transmission capacity that can be built between any two regions, by time. This is equal to 1GW by 2020, 4GW by 2025, 8GW by 2030, 16GW by 2035, and unlimited (but totals still subject to the national constraint) thereafter.
The EIA State Energy Data System (EIA November 2015) provided data on historical electric energy flows between states and regions, and data on international exports and imports, in addition to regional prices. These data were used in the model calibration process. In the current version of the model, international imports and exports of electricity are assumed to hold constant over time, at the most recently available historical levels determined from SEDS (2015). This includes electricity traded with Canada and Mexico. Figure 2‑5 shows the 2015 net trade positions (after accounting for transmission and distribution losses) for the default 16 regions in US-REGEN.
New wind and solar capacity incurs a one-time charge of $250 per kW and $100 per kW respectively to reflect the incremental transmission investment required to incorporate generally remotely located wind and solar capacity within a region. This number is based on a combination of discussions with internal and external transmission experts and a review of the relevant literature (for example see Mills et al. (2009), Jaske (2010)). There is no added transmission charge for repowered wind (i.e. new wind turbines replacing existing turbines at existing sites). An incremental distribution charge for rooftop PV is included as described in Section 3.6.
With the exception of these specific cost adders for transmission inter-connection costs of wind and solar, the model does not explicitly represent transmission or distribution within a region. As described in Section 2, it does account for the expenditures on these components as a component of the retail price.
2.3.7 Emissions
SO2 and NOx rates for existing units are constructed from two sources: unit-level EPA Continuous Emission Monitoring System (CEMS) data, as provided by ABB Energy Velocity, and EIA plant-level data for those units not reported in the CEMS dataset, typically small units. These emission rates are aggregated to the US-REGEN blocks using a capacity-weighted average. As these data are sourced from 2017 and 2018, not all the emission rates will reflect the addition of environmental controls to meet EPA regulations. Therefore, for the 2020 model time step and beyond, the emission rates are adjusted to ensure all units comply with the Mercury and Air Toxics Standards (MATS). SO2 rates are set equal to the minimum of the observed rate and the 0.2 lb/mmbtu MATS standard. NOx rates for existing units are held constant at the observed rate in 2018. REGEN includes a representation of the Cross-state Air Pollution Rule for both the annual fine particulate and the ozone season NOx requirements, but these constraints do not bind given assumed emission rates.
Emission rates for new units are based on data from EPRI's Generation Options report and historic emission rates from modern units and meet all current regulations for SO2 and NOx emission rates. New supercritical pulverized coal units without CCS are assumed to emit 0.075lb NOx/mmbtu, and 0.0858 lb SO2/mmbtu. New IGCC coal units with CCS are assumed to emit 0.051 lb NOx/mmbtu and 0.0017lb SO2/mmbtu, based on the best 12% of existing coal units. New combined cycle gas-fired units are assumed to emit 0.009 lb NOx/mmbtu.
CO2 emission rates are calculated from the average carbon content of the input fuel, multiplied by the average heat-rate of the unit in question. The carbon content of coal, refined petroleum, and natural gas respectively is set at 0.093, 0.073, and 0.054 tons CO2 per mmbtu. For units with carbon capture and storage, this rate is reduced by the assumed CO2 capture percentage.
This percentage was determined based on input from the EPRI Nuclear Sector's long-term operation experts. It can be varied in certain scenarios, for example in some cases all licenses are assumed to end at 60 years. No additional cost is incurred to extend the unit's life. ↩︎
There was no existing offshore wind capacity at the end of 2015. ↩︎
CCS = carbon capture and storage. ↩︎
It should be noted that projecting future technology costs for use in the US-REGEN model is an ongoing, evolving process. Costs are updated regularly, particularly costs for wind and solar. ↩︎
See https://www.netl.doe.gov/coal/carbon-storage/strategic-program-support/natcarb-atlasopen in new window. This database is populated from the Department of Energy's Carbon Storage Atlas – Fifth Edition (Atlas V), also available via this link. ↩︎
Both coal and gas can be marginal generators depending on the region and time of day. Even in 2007, when gas prices were very high, not all coal units were in the money all the time. ↩︎
This approach is also known as a pipeline model, albeit with a very high level of aggregation. It does not take into account the network effects inherent in Kirchoff's laws. ↩︎