Rensselaer Polytechnic Institute & Cornell University, 2 Cornell University, 3. Arizona State University, 4 US Federal Energy Regulatory Commission

Transcription

1 Estimating the Long-Run Effects of Environmental Policies on the Electricity Grid: Prices, Investment, Demand Response and Resulting Carbon Dioxide Emissions Daniel Shawhan 1, John Taber 4, William Schulze 2, Ray Zimmerman 2, Daniel Tylavsky 3, Di Shi 3, Douglas Mitarotonda, Charles Marquet 2, and Jubo Yan 2 1 Rensselaer Polytechnic Institute & Cornell University, 2 Cornell University, 3 Arizona State University, 4 US Federal Energy Regulatory Commission

2 Disclaimer for Author Who Now Works at US Federal Energy Regulatory Commission The views expressed in this paper do not necessarily represent the views of the Federal Energy Regulatory Commission or of the United States government.

3 Overview Proper assessment of electricity policies, potential new generation units, and potential expansions of the power transmission system requires prediction of their system-wide, society-wide, and long-term effects. We present a model designed with this purpose in mind. We provide a preliminary demonstration of its use, to assess a few potential US sustainable energy policies.

4 Inadequate Attention to Electric Grid Characteristics Can Lead to Errors ~$6 billion lost due to 8/14/03 blackout

5 Power Flows Over All Lines According to Kirchoff s Voltage Law, and Our Model Reflects This If each segment has equal resistance and length, Kirchoff s Law states that 1/3 of the power will flow along the upper path and 2/3 will flow along the bottom path, regardless of how close each line is to its safe transmission limit.

6 Part 1: The SuperOPF Planning Tool

7 SuperOPF Planning Tool Determines flows according to laws of physics Predicts or optimizes system operation, investments, retirements, and effects. Includes emissions and estimated health damage. Can be used with model of any grid. We are developing US-Canadian models. Will be publicly available and modifiable.

8 More Detail on Features of SuperOPF Planning Tool Optimizes operation, investment, and retirement every ten years: Minimizes sum of operating, investment, and demand response costs over the next ten years 12 representative hours (soon, 75) Load responds to price

9 Environmental Features Calculates emissions of CO2, NOX and SO2 (so far) Uses air pollution transfer coefficients from EPA-funded model to estimate effect of each generation unit (per MWh) on the fine particulate matter in each county of the US We are adding transfer coefficients for Canada These allow estimation of mortality and illness effects of any policy or investment

10 Verification of Air Quality Aspect of Modeling Annual 2010 US premature mortality estimated from our data on the generators is 13,000, which is entirely consistent with a prior study s national estimate using the same exposure-response function. Health damages and mortality not included in example runs shown today

11 Network Models We Have Developed We use three network reductions to cover the entire nation. These reductions retain all high voltage lines of 230 KV and congestion-prone lower-voltage lines. Equivalent lines represent those not retained node equivalent model for East (EI) node equivalent model for West (WECC) node equivalent model for Texas (ERCOT)

12 We Use a Reduced Model That Retains All High-Voltage Lines and Represents the Rest with Aggregated Equivalent Lines

13 Comparison of Optimal Power Flow Simulation Results: Simplified vs. Full Models of Eastern Interconnection Original Model Simplified Model Nodes ~62,000 ~5,200 Convergence of the solution (Y/N) Time for Convergence (sec) Y Y Total Cost ($/Hour) 16,244,322 16,193,495 Average Price ($/MWh)

14 Selected Generator and Load Data Detailed information about existing generation units combined from 12 sources Investment costs from US EIA Fuel cost projections from US EIA Twelve hour types represent the year. Vary in terms of unit availability (from NERC) and load (from ISOs and NERC). Load grows at 0.59% per year before long run demand response * * _(2).pdf

15 How the Model Makes Its Predictions It finds the combination of plant construction, retirement, and operation that maximizes Consumer benefits - Construction costs - Annual fixed costs - Operating costs over the decade, subject to meeting load and respecting network constraints.

16 Typical Run Adjust input parameters to reflect a policy, investment, or behavior Sequential optimization of three periods 2012 current fleet 2022 allowing retirement and new investment 2032 allowing retirement and new investment

17 Problem Size For 5222-bus Eastern Interconnect model, with 2882 aggregated generators, the simulation is a linear programming problem with 750,000 variables, 2,000,000 constraints Solves in about an hour on a 12-processor workstation with 64 GB of RAM

18 Part 2: An Example of Simulation Results in the Eastern Interconnection (75% of US plus part of Canada)

19 Investment and Retirement Base year is Investment allowed in 2022 and New plants must recoup capital in 10 years. Existing plants that cannot cover their annual fixed costs (for operating, maintenance, tax, and insurance) are retired.

20 New Power Plant Costs Fuel Type Capital Recovery Required ($/MW/Year) Annual Total Fixed Costs ($/MW) Total Variable Cost $/MWh Total Possible Capacity Additions in East Coal (Dual Unit $497,201 $35,255 $ GW Advanced PC) Natural Gas $181,824 $20,661 $39.05 (if $ GW (Advanced NGCC) per Mcf; varies) Wind* $392,322 $30,710 $0 249 GW (2022) 285 GW (2032) Nuclear w/ subsidies $470,226 $95,571 $ GW & loan guarantees Solar* $520,000 (2022) $390,000 (2032) $17,548 $0 250 GW (2022) 285 GW (2032) *Excluding tax credits for wind and solar (included in some runs). To calculate generation from capacity, we use six solar zones and six wind zones, each with a zone-specific capacity factor. Sources: U.S. EIA, Updated Capital Cost Estimates for Electricity Generation Plants, Nov and U.S. EIA, Annual Energy Outlook 2011, April 26, Except solar capital recovery required, which is from those sources but multiplied by 0.68 for 2022 and 0.51 for 2032.

21 Cases 1) Base HG : Base Case, High Gas Price 2) Base LG : Base Case, Low Gas Price 3) C&T HG : Cap and Trade, Wind & Solar Tax Credits, High Gas Price 4) C&T LG : Cap and Trade, Wind & Solar Tax Credits, Low Gas Price 5) EPA HG : No new coal units, Wind & Solar Tax Credits, High Gas Price 6) EPA LG : No new coal units, Wind & Solar Tax Credits, Low Gas Price

22 High Natural Gas prices assume low prices in the short term due to shale gas, which increases in 2022 due to depletion and converges to the world price by Building Blocks of Cases: Natural Gas Prices Low Natural Gas prices are estimates from the EIA 2011 Coal and Oil costs are assumed to remain unchanged $/MBTU Natural Gas $2.50 $7 $14 (High) Natural Gas (EIA) $2.50 $4.77 $5.86

23 Building Blocks of Cases: CO 2 Cap and Trade Program Based on Kerry-Lieberman legislation, which included a ceiling and a floor on the price of the emission permits. Assume price is at the ceiling, which increases by 5% annually $36.94/metric tonne in 2022 $60.18/metric tonne in 2032

24 Building Blocks of Cases: Proposed USEPA CO 2 Regulation 1000 pound per MWh limit for new baseload and shoulder fossil fueled plants Coal-fired plants can comply only if adopt carbon capture & sequestration within ten years We assume no new coal-fired units will be built as a result

25 Building Blocks of Cases: Tax Credits for Wind and Solar $22/MWh for the first ten years Represents federal Production Tax Credit for wind and Investment Tax Credit for solar

26 Cases 1) Base HG : Base Case, High Gas Price 2) Base LG : Base Case, Low Gas Price 3) C&T HG : Cap and Trade, Wind & Solar Tax Credits, High Gas Price 4) C&T LG : Cap and Trade, Wind & Solar Tax Credits, Low Gas Price 5) EPA HG : No new coal units, Wind & Solar Tax Credits, High Gas Price 6) EPA LG : No new coal units, Wind & Solar Tax Credits, Low Gas Price

27 Wind Additions GW Base HG Base LG C&T HG C&T LG EPA HG EPA LG The largest share of wind is built in the Southwest Power Pool (mostly Kansas and Oklahoma) Some wind is built in the other regions (except for Florida)

28 Solar Additions GW Base HG Base LG C&T HG C&T LG EPA HG EPA LG Solar is built in Florida in all cases. In Kerry-Lieberman and EPA cases, which include the wind & solar tax credits, solar is also built in the southeast US

29 Nuclear Additions GW Base HG Base LG C&T HG C&T LG EPA HG EPA LG Only in the Kerry- Lieberman High Gas case is Nuclear built. Limits for Solar and Wind reached, Nuclear cheaper than NGCC at $14 gas

30 Coal Retirements 0 GW Base HG Base LG C&T HG C&T LG EPA HG CO2 emissions charges force greater retirement of coal units; wind & solar tax credits allow more units to be retired in favor of renewables -250 EPA LG

31 NG Additions and Retirements NGCC is built in New England in all cases GW Base HG Base LG C&T HG C&T LG EPA HG In the Base Case, some NGCC is built in Florida under both gas prices and PJM under low gas prices. In Cap & Trade, NGCC is built in all regions EPA LG Retirements are much larger than additions as older, inefficient units are retired -200

32 Generation from Coal TWh per year 1,800 1,600 1,400 1,200 1, Base HG Base LG C&T HG C&T LG EPA HG EPA LG Increases Substantially Under Base Cases Because of Higher Gas Price and Increased Load (blue) Increases Somewhat with Wind & Solar Tax Credits and Ban on New Coal-Fired Generators (red) Declines Dramatically with Cap & Trade + Wind & Solar Tax Credits (green)

33 Results: CO2 Emissions CO2 (Megatonnes) 2,500 2,000 1,500 1, Base HG Base LG C&T HG C&T LG EPA HG EPA LG EPA lowers CO2 not by eliminating coal but because of wind & solar tax credits In the Base Cases, more older NG are eliminated in the HG case, which lowers CO2 emissions in 2032, though less fuel switching occurs In the C & T case, higher gas prices result in less Coal->NG fuel switching, which increases CO2 emissions

34 Results: Average Wholesale Prices 120 $/MWh Base HG Base LG C&T HG C&T LG EPA HG EPA LG Cap and Trade would increase average electricity prices substantially. EPA cases lower prices compared to base cases because of tax credits for wind & solar

35 Results: Total Economic Surplus (Ignoring Environmental Damages) Consumer + Producer + Government Economic Surplus in $Billions per Year Base HG Base LG C&T HG C&T LG EPA HG EPA LG In the initial decade of new plant construction (2022), the environmental and renewable energy policies reduce this measure of social welfare because of higher wind, solar, and gas plant construction. In the following decade (2032), the same policies increase this measure of welfare because of fuel savings associated with more reliance on renewable energy.

36 Conclusions of Sample Simulation CO 2 CO 2 will increase without additional regulations Tax credits for wind & solar reduce CO 2 ~25% Cap & trade + tax credits for wind & solar reduce CO 2 ~75% Generation Even without EPA CO 2 limit on new generators, no coal built. Even without wind & solar tax credits or environmental regulations, a range of generation technology (NGCC, wind and solar) is built. Wind and solar driven to build limits if tax credits Nuclear built only in Cap & Trade High Gas Price case

37 RGGI Cap & Trade on Power Plant CO 2 in NE USA: Leakage to Neighboring States & Provinces

38 Cost vs Emission Reductions Under Five Policy Scenarios

39 Integration of Secondary Fine Particulate Mortality Cost Into Model Damage cost per MWh at each generator calculated from Unit-specific emission rates from EPA COBRA-based SO2-PM2.5 and NOX-PM2.5 transfer coefficients (so far, for US only) Dose-response function from Abt Associates Projected county population data for 2010 US averages by fuel type applied to Canadian generators (until can add Canadian transfer coefficients)

40 Mortality Co-Benefits of $50 CO2 Price Are More Than Twice the Costs Costs ($ Millions per Year) Fuel + VOM 2dary PM2.5 mortality Sum Policy A: Status quo (SO2 $700, NOX $2000) $13,929 $19,896 $33,825 Policy B: $50 CO2 price (SO2 $700, NOX $2000, CO2 $50) $14,103 $19,481 $33,584 Change as a result of $50 CO2 price $173 -$415 -$241

41 Conclusions About Modeling Method Can predict effects of policy combinations. Can predict effects of generator or transmission system change. Electrical model that can predict operation, investment, retirement, prices, economic surplus, emissions, and health effects of modeled scenario. Can optimize investment. Can be adapted to any region.

42 END

43 CONTINGENCY SLIDES

44 Converting pollution into estimated mortality cost Seventy million county-to-county transfer coefficients from EPA-funded model Population per county, and percentage over 30, from US and Canadian censuses Exposure-response functions from NRC Valuation per premature death is US EPA standard value

45 Representative hours 12 hours represent the year 4 each for summer, winter, and fall/spring For example, summer peak hour type represents the 5% of summer hours with the highest hourly Eastern load Load varies independently in each region of the East, based on actual 2010 data for each region and representative hour bin

46 The SuperOPF Planning Tool: Already has piecewise linear benefits (step function long run demand response of load to price) DC load flows Allows optimal investment in new generation by type Uses a detailed network reduction that retains all high voltage lines (e.g., 5200 node reduction of Eastern Interconnection) Add optimization of capacity of selected additional new and existing lines Could be solved as a quadratically constrained mixed integer optimization problem for which good solvers exist

47 Long Run Price Response 25% total demand response 10 blocks, each 2.5% of load Effective price is at the midpoint of each interval Consistent with piecewiselinear benefits function (Conservation) Actual Demand Response (Red) Modeled Demand Response (Blue) LMP ($/MWh) Load (Percent of Base) In the long run, the elasticity of demand for electricity is - 1. (1) We use an elasticity of 0.8 to combine the short and long run. The delivered price equals the LMP for each bus, for each representative hour in 2012, plus estimated distribution costs ($70/MWh) 1. Dahl, Carol. A Survey of Energy Demand Elasticities in Support of the Development of the NEMS. ftp://ftp.eia.doe.gov/pub/oiaf/elasticitysurvey/elasticitysurvey_dahl.pdf

48 Future Work: Example Proposals for New or Upgraded Transmission Proposal in New York State for a new line under the Hudson river to connect Hydro Quebec to NYC backup for wind BPA Proposal to upgrade I-5 corridor to provide power to Portland Oregon shortages expected to develop in a few years Texas needs new lines for anticipated wind expansion

49 For each US generation unit, we match up the data from the following data sets: EIA EGU list (master list, state, max MW, fuel type, unit type) EIA plants list (latitude and longitude, which NERC region unit is in) Energy Visuals (MMWG) Transmission Atlas EGU list (location on network) EPA Continuous Emission Monitoring hourly data (heat rate) EPA Clean Air Markets EGU list (emission rates) Energy Visuals FirstRate EGU list (heat rate if not available from EPA) EIA flue list (stack(s) associated with each unit, and stack parameters necessary to calculate effective stack height of each) Air pollution transfer coefficients for each unit from EPA contractor, same as used in National Research Council s Hidden Costs of Energy study

50 The match-ups that require experimentation and sophisticated coding EPA unit list with other datasets EIA units with Energy Visuals units (those not already matched) Combined cycle parts with each other Units with interconnections (because of errors in data on which NERC region each unit is in) Units not in Energy Visuals data with network nodes EIA flues with EIA units Air pollution transfer coefficients with other datasets

51 Representative Hours in EI Model Relative Frequency Average Load 18% 100% 16% 14% 12% 10% 8% 6% 4% 2% Peak High Medium 80% 60% 40% 20% 0% Summer Winter Fall+Spring Low 0% Summer Winter Fall+Spring

52 Most sophisticated data processing aside from the matchups Heat rate of each unit from EPA CEMS hourly data only the hours during which units are at or above 70% of max capacity Emission functions of each unit from EPA CEMS hourly data (in progress), taking into account effects of ramping and start-up Ramp rate of each unit from EPA CEMS hourly data (future) Aggregation of similar units to reduce number of control variables

53 Missing data we are filling in with predicted values Heat rates, VOM, and emission rates using averages by subtype or regression analysis At units for which we have full info, calculate average by subtype or average relationship between universally known variables (e.g. age, size, pollution controls) and the ones above Use those relationships to estimate the unknown variables at units where they are missing Use non-linear regression analysis in polar coordinates, and atmospheric science principles, to estimate missing pollution transfer coefficients

54 Problems with MMWG data for which we are correcting Phantom units (in MMWG data) Incorrect heat rates (in supplementary data)

55 Features of SuperOPF Planning Tool, 1 Can work with detailed electrical grid models Can simulate alternating-current operation of grid, but we use it in direct-current linear approximation mode because that is faster and produces very similar results. Also, many of the differences would be ephemeral.

56 Demand Response to Price Demand response can be modeled at each consumption location as blocks of generation, with successively higher prices Elec. Price ($/MWh) Estimated Actual Demand Response (Red) Modeled Demand Response (Blue) Load (Percent of Base)