Integrating Traditional, Variable, Renewable, Distributed, and Demand-Side Response Resources G RID S CHOOL 2010 M ARCH 8-12, 2010  R ICHMOND, V IRGINIA I NSTITUTE OF P UBLIC U TILITIES A RGONNE N ATIONAL L ABORATORY Thomas D. Veselka Center for Energy, Economic, and Environmental Systems Analysis Decision and Information Sciences Division ARGONNE NATIONAL LABORATORY  Do not cite or distribute without permission MICHIGAN STATE UNIVERSITY

Veselka - 02 GridSchool 2010 The Grand Challenge of Integrating Renewable Resources with Variable and Intermittent Production into the Grid Is the Ability to Respond to Rapid and Unpredictable Fluctuations Day of the Month in August Wind Output (MW) Total of 3 Sites Wind is not Always Available when Needed Most Rapid Ramping The thermal system or loads need to adjust quickly

Veselka - 03 GridSchool 2010 Wind Probability Profiles Vary Seasonally and by Time of Day Exceedance Probability (%) Wind Production (MW) Summer Nighttime Wind Is Less than Daytime Wind Winter Wind Is Greater Than Summer Wind On Average, Wind Output Decreases in the Morning When Load Is Rapidly Increasing. The Opposite Occurs in the Evening.

Veselka - 04 GridSchool 2010 Wind Resources Vary Widely Across the United States Often the Best Wind Resources Are Far from Major Load Centers Transmission MISO PJM 765 kV

Veselka - 05 GridSchool 2010 The U.S. has Installed the Most Wind Capacity in the World, but the Percent Penetration Rate ( % Production ) Is Relatively Small U.S. recently became the world leader in wind power with over 8 GW installed in 2008 and 25 GW total installed capacity (AWEA, Feb 09)

Veselka - 06 GridSchool 2010 Wind Capacity by State

Veselka - 07 GridSchool 2010 U.S. Wind Capacity Growth Source: AWEA 2009

Veselka - 08 GridSchool 2010 Does Wind Power Influence Market Operations? Negative prices (LMPs) Wind power ramping events Midwest ISO Wind Power and MN Hub Prices, May 11-17, 2009:

Veselka - 09 GridSchool 2010 United States Photovoltaic Solar Resource Map

Veselka GridSchool 2010 There Are Several Different Types of Photovoltaic Technologies, Each of Which Has its Own Set of Attributes One Size Does Not Fit All Luminescent Solar Concentrators with Multijunction Cells (~40%)

Veselka GridSchool 2010 Photovoltaic Efficiencies Have Increased Dramatically Since the Mid-1970’s and Are Expected to Continue Improve Source:

Veselka GridSchool 2010 Public Service Company of Colorado Solar Study 200 MW of PV & 200 MW CSP with 800 MWh Storage Source:

Veselka GridSchool 2010 Spanish PV Study: Annual Hourly Photovoltaic Output Generation (kW) Source:

Veselka GridSchool 2010 Most States Have a Renewable Portfolio Standard (RPS) Source: / September 2009 State renewable portfolio standard State renewable portfolio goal Solar water heating eligible * † Extra credit for solar or customer-sited renewables Includes separate tier of non-renewable alternative resources WA: 15% by 2020* CA: 20% by 2010 ☼ NV : 25% by 2025* ☼ AZ: 15% by 2025 ☼ NM: 20% by 2020 (IOUs) 10% by 2020 (co-ops) HI: 40% by 2030 ☼ Minimum solar or customer-sited requirement TX: 5,880 MW by 2015 UT: 20% by 2025* ☼ CO: 20% by 2020 (IOUs) 10% by 2020 (co-ops & large munis)* MT: 15% by 2015 ND: 10% by 2015 SD: 10% by 2015 IA: 105 MW MN: 25% by 2025 (Xcel: 30% by 2020) ☼ MO: 15 % by 2021 WI : Varies by utility; 10% by 2015 goal MI: 10% + 1,100 MW by 2015* ☼ OH : 25% by 2025 † ME: 30% by 2000 New RE: 10% by 2017 ☼ NH: 23.8% by 2025 ☼ MA: 15% by % annual increase (Class I Renewables) RI: 16% by 2020 CT: 23% by 2020 ☼ NY: 24% by 2013 ☼ NJ: 22.5% by 2021 ☼ PA: 18% by 2020 † ☼ MD: 20% by 2022 ☼ DE: 20% by 2019* ☼ DC: 20% by 2020 VA: 15% by 2025* ☼ NC : 12.5% by 2021 (IOUs) 10% by 2018 (co-ops & munis) VT: (1) RE meets any increase in retail sales by 2012; (2) 20% RE & CHP by states & DC have an RPS 5 states have goals KS: 20% by 2020 ☼ OR : 25% by 2025 (large utilities )* 5% - 10% by 2025 (smaller utilities) ☼ IL: 25% by 2025 Standards Should be Consistent with Renewable Resources & Needs

Veselka GridSchool 2010 Currently, Volatility in Production from Variable Resources Are Accommodated by Changing Thermal Unit and Hydroelectric Power Plant Production Levels Time Output (MW) Operating Capacity Ramp Up Rate Ramp Down Rate Min Output Cold Start Time Minimum Down Time Minimum Up Time Load Following Range The Greater the Operational Flexibility of Dispatchable Units, the more Variability the Grid Will Accommodate

Veselka GridSchool 2010 Some Technologies Are Able to Come On-line Quickly to Respond to Rapid Load Changes while Others Respond More Slowly Weeks for Shutdown and Startup Some Hydropower Plants Change Very Quickly

Veselka GridSchool 2010 The Load Following Range Is Restricted by the Output Minimum and Generation Capacity Time Production (MW) Operating Capacity Ramp Up Rate Ramp Down Rate Cold Start Time Minimum Down Time Minimum Output Load Following Range Minimum Up Time

Veselka GridSchool 2010 GT Highest Production Costs NGCC Cycling Coal Base Load Coal Ideally, Units Are Dispatched Based on Production Cost Hour of the Day Load (MW) Max Load Min Load Nuclear Lowest Production Costs Nuclear 12 $/MWh Coal 25 $/MWh NGCC 41 $/MWh Cycling Coal 32 $/MWh Gas Turbines 80 $/MWh Resource Stack Supply (MW)

Veselka GridSchool 2010 Unfortunately, a Steam Plant ( e.g., Cycling Coal ) Does not Have the Flexibility to Operate at a Very Low Output Level GT Highest Production Costs NGCC Cycling Coal Base Load Coal Hour of the Day Load (MW) Max Load Min Load Nuclear Lowest Production Costs Ramp Up Ramp Down GT Operations

Veselka GridSchool 2010 Wind Production Will Serve Some of the Load. This Production Reduces the Loads that Are Served by other Generating Resources Hour of the Day Load/Wind Output (MW) Max Load Min Load Wind Generation Ramp Up Ramp Down

Veselka GridSchool 2010 Dispatchable Units Serve a Load Profile that Typically, but not Always, Has Greater Fluctuations Relative to the Case where there Is no Wind Hour of the Day Load (MW) New Max New Min Wind Typically Increases Resultant Load Changes Larger Range of Operations Ramp Up Ramp Down

Veselka GridSchool 2010 NGCC GT Cycling Coal Nuclear Base Load Coal Lowest O&M Costs Highest O&M Costs Unit Dispatch with Wind Results in Less Thermal Generation & Associated Air Emissions Hour of the Day Load (MW) Without Wind With Wind Coal May Operate Less Min Gen

Veselka GridSchool 2010 Wind and other Renewable Technologies Will Reduce Greenhouse Gas Emissions 20 Percent Wind by 2030 Report: CO 2 Emissions Are Estimated at 25 Percent Lower Than a No-Wind Scenario

Veselka GridSchool 2010 As a Result of Variable Resource Generation Some Units Will Operate at a Different Efficiency Point Combined Cycle Hydro Diesel Nuclear Fossil Steam

Veselka GridSchool 2010 NGCC GT Cycling Coal Nuclear Base Load Coal Lowest O&M Costs Highest O&M Costs Unit Dispatch with Greater Nighttime Wind Base Load Coal Unit May Need to Be Taken Off-line for Several Hours Hour of the Day Load (MW) Without Wind With Wind Sell Sell

Veselka GridSchool 2010 Pump Energy is Produced When Generating Energy is Consumed When Pumping Substation Upper Reservoir Lower Reservoir Pumped Storage Plants Can Be Used to Help Smooth Out Loads Served by Other Dispatchable Resources Fill Load Valley (Consume) to Utilize Low Cost Production and Avoid Expensive Shutdown Costs React to Sudden Changes in Variable Resource Production Load (MW) Hour of the Day

Veselka GridSchool 2010 When a Unit Is Forced Out of Service, the System Responds by Altering the Dispatch Nuclear 8 $/MWh Base Coal 25 $/MWh ,000 1,250 1,500 1,750 2,500 2,250 2,000 Supply Marginal Cost 1,200W 40$/MWh Supply Stack without Maintenance (MW) Cycling Coal 40 $/MWh Gas Turbines 80 $/MWh NGCC 60 $/MWh ,000 1,250 1,500 1,750 2,500 2,250 2,000 Supply Marginal Cost 1,200W 80$/MWh Supply - Nuclear Unit Out of Service (MW) Cycling Coal 40 $/MWh Coal Steam Partially Loaded Base Coal 25 $/MWh Unprepared Slow Transition Operational Problems After the outage it will take hours for the system to reach the least-cost state of operations All demand will not be served Least-Cost Resource Stack Before Outage Least-Cost Resource Stack After Outage

Veselka GridSchool 2010 Spinning Reserves Help Alleviate Operational Problems Associated with Random Outages Nuclear 8 $/MWh Base Coal 25 $/MWh ,000 1,250 1,500 1,750 2,500 2,250 2,000 Supply Marginal Cost Supply Stack without Outages (MW) Cycling Coal 40 $/MWh Gas Turbines 80 $/MWh NGCC 60 $/MWh 1,200 MW 80$/MWh ,000 1,250 1,500 1,750 2,500 2,250 2,000 Supply Supply - Nuclear Unit Out of Service (MW) Cycling Coal 40 $/MWh Base Coal 25 $/MWh NGCC 60 $/MWh Gas Turbines 1,200 MW Load 80 $/MWh Fast Transition Simultaneously Ramp Operations Spinning Reserves NGCC 250 MW NG Steam 110 MW Oil Steam 110 MW Gas Turbines 30 MW Total 500 MW

Veselka GridSchool 2010 System Operators Need to Make Certain that Ramping Resources Are Available Nuclear 8 $/MWh Base Coal 25 $/MWh ,000 1,250 1,500 1,750 2,500 2,250 2,000 Supply Supply Stack without Maintenance (MW) Cycling Coal 40 $/MWh NGCC 60 $/MWh Gas Turbines 6 AM Load 1200 MW 7 AM Load 1500 MW ,000 1,250 1,500 1,750 2,500 2,250 2,000 Supply Nuclear 8 $/MWh Base Coal 25 $/MWh Cycling Coal 40 $/MWh NGCC 60 $/MWh Gas Turbines Old GT 120 $/MWh Supply ,000 1,250 1,500 1,750 2,500 2,250 2,000 Nuclear 8 $/MWh Base Coal 25 $/MWh Cycling Coal 40 $/MWh 8 AM Load 1800 MW Spinning Reserves 500 MW Spinning Reserves 500 MW Spinning Reserves 500 MW Spinning Reserves 500 MW Spinning Reserves 500 MW Spinning Reserves 500 MW NGCC 60 $/MWh Gas Turbines Old GT 120 $/MWh

Veselka GridSchool 2010 Usually, the Grid Can Accommodate Relatively Small Amounts of Wind Generation Nuclear 8 $/MWh Base Coal 25 $/MWh ,000 1,250 1,500 1,750 2,500 2,250 2,000 Supply Supply Stack without Maintenance (MW) Cycling Coal 40 $/MWh NGCC 60 $/MWh 6 AM Load 1200 MW 7 AM Load 1500 MW ,000 1,250 1,500 1,750 2,500 2,250 2,000 Supply Nuclear 8 $/MWh Base Coal 25 $/MWh Cycling Coal 40 $/MWh NGCC 60 $/MWh Gas Turbines Old GT 120 $/MWh Supply ,000 1,250 1,500 1,750 2,500 2,250 2,000 Nuclear 8 $/MWh Base Coal 25 $/MWh Cycling Coal 40 $/MWh 8 AM Load 1800 MW Spinning Reserves 500 MW Spinning Reserves 500 MW Spinning Reserves 500 MW Spinning Reserves 500 MW Spinning Reserves 500 MW Spinning Reserves 500 MW NGCC 60 $/MWh Ramp up due to wind decrease WIND

Veselka GridSchool 2010 The Unpredictability of Wind Compounds Grid Integration Problems Forecast of Wind Power Production Levels Can Be Made for the Next Few Days Source: Iberdrola Renewables, 2009 Eyeballing: Looks pretty good Mean absolute error is 9.3% But devil is in the details (ramps)

Veselka GridSchool 2010 Wind Forecasts Are Far from Perfect in the Short-Term and Much Worse in the Long-Term Error depends on several factors – Prediction horizon – Time of the year – Terrain complexity – Model inputs and model types – Spatial smoothing effect – Level of predicted power Errors in SCADA information and wind farm operation Error in meteorological forecasts Errors in wind-to- power conversion process Phase Error Magnitude Error

Veselka GridSchool 2010 Technology Improvements Are Alleviating Some Problems Example: The Danish Horns Rev Wind Farm Is Providing Regulation (Frequency Response) and Balancing Response Source: Smith et al., IEEE Power and Energy Magazine, Vol. 7. No.2, Control Wind Output with Blade Pitch (Spill Energy)

Veselka GridSchool 2010 Historical Winter Load Shapes and Wind Generation in the Midwest Source: Wind ~ 14% of Load Unit Commitment Study

Veselka GridSchool 2010 Problem: Given that Wind Forecasts Have Errors, Make an Economic Unit-Commitment Schedule that Is Reliable Source: Costs: Generation, Unserved Energy, & Startup Constraint: Ramping, Up & Down Time, Min Output, & Maintain Reserves Wind can be curtailed (spilled energy)

Veselka GridSchool 2010 Unit Commitment Results Using Various Modeling Methodologies and Assumptions Source: Stochastic UC gives higher commitment & more available operating reserves Similar result for deterministic UC w/additional reserve requirement

Veselka GridSchool 2010 Comparison of Costs (30 day simulation) Results Based on Fixed Unit Commitments and Real-Time Economic Dispatch  The potential value of forecasting illustrated by perfect forecast (D1)  Deterministic UC with point forecast (D2) appears too risky  Deterministic UC w/add reserves (D3) and stochastic UC (S1) give similar total cost Source:

Veselka GridSchool 2010 Finding the “Best” Mix of Generating Capacity to Backup Variable Resources While Keeping Costs Reasonable Is Challenging TechnologyConstruction CostOperating Cost*Operating Flexibility** Fossil Steam242 Hydroelectric234 Combined Cycle333 Gas Turbine525 Diesel Generator415 Nuclear Steam151 Desirability Rating 1Very Low 2Moderately Low 3Average 4Moderately High 5Very High * Operating costs includes fuel costs and fixed and variable operating and maintenance costs ** Operating flexibility is the unit’s ability to respond to load changes and includes ramp rates, cold start time, etc. Variability Issues Zero Fuel Costs

Veselka GridSchool 2010 Suggested Reading: DOE’s 20% Wind by 2030 Report  Explores “ a modeled energy scenario in which wind provides 20% of U.S. electricity by 2030 ”  Describes opportunities and challenges in several areas  Turbine technology  Manufacturing, materials, and jobs  Transmission and integration  Siting and environmental effects  Markets  Enhanced wind forecasting and better integration into system operation is one of the challenges  DOE is funding several research projects in this area

Thank you for your attention Source: BOR

EXTRA SLIDES Maximum Variable Resource Capacity Credit

Veselka GridSchool 2010 Reserve Capacity Is Needed to Serve Load when One or More Generators Are Out of Service Peak Load Forecast Years MW Upper RM Lower RM Total System Capacity Existing System Capacity New Capacity Additions Engineering Guideline Build 15% to 20% more capacity than the peak load

Veselka GridSchool 2010 Variable Resources Have a Capacity Value Which Can Be Approximated Using Probabilistic Methodologies Probability of 3 sixes = 1/6 x 1/6 x 1/6 = 1/216 = less than %0.5

Veselka GridSchool 2010 Loads Can Also Be Viewed as Probabilistic Events Step 1: Chronological Loads Hour of the Day Load (MW) Max Load Min Load

Veselka GridSchool 2010 Sorted Summer Load Profile Step 2: Load Exceedance Curve hr 4 hr 17 Load (MW) Max Load Min Load Time Load Some Information Is Lost Such as Load Changes Over Time hr 17 hr Exceedance Probability (%)

Veselka GridSchool 2010 NGCCGT Cycling Coal Base Load Coal Nuclear Unit Production Levels Can Be Estimated Using a Load Duration Curve 0100 Exceedance Probability (%) Load (MW) Max Load Is Never Exceeded Time Load Min Load Is Always Exceeded Information Such as Unit Ramping and Frequency of Unit Starts/Stops Are Lost

Veselka GridSchool 2010 NGCC GT Cycling Coal Base Load Coal When a Supply Resource Is Forced Out of Service Other Resources Are Dispatched to Serve the Load Hour of the Day Load (MW) Max Load Min Load Nuclear Forced Out of Service Time Load

Veselka GridSchool 2010 NGCC GT Cycling Coal Base Load Coal Alternatively, the Load Curve Can Be Adjusted While Including the Out-of-service Unit Hour of the Day Load (MW) New Max New Min Nuclear Original Curve Load not Served by the Nuclear Unit Is Satisfied by Other Units in the Resource Stack Nuclear Capacity Time Load

Veselka GridSchool 2010 NGCC GT Cycling Coal Base Load Coal The Same Methodology Can Be Applied to a Load Duration Curve 0100 Exceedance Probability (%) Load (MW) Nuclear New Max New Min Nuclear Capacity Original Curve

NGCC GT Cycling Coal Base Load Coal There Is Some Probability that a Unit Does Not Operate 0100 Exceedance Probability (%) Load (MW) We don’t know with certainty if the nuclear unit will be either on or off at some point in the future Nuclear Capacity Equivalent Load Curve Accounts for Nuclear Outages Nuclear Weighted Average Curve of Nuclear Unit On & Off Nuclear Off Nuclear On Area = Outage Rate/100 X Nuclear Capacity

Likewise All Units Are “Convolved” Into the Load Duration Curve Exceedance Probability (%) Nuclear NGCC GT Cycling Coal Base Load Coal 0100 Load (MW) Operating System Capacity Nuclear Cycling Coal Base Coal GT NGCC Energy Not Served Original Curve Total Capacity + Peak Load Final Equivalent Load Curve Accounting For All Unit Outages Loss of Load Probability

Exceedance Probability (%) Load (MW) Without Wind With Wind Using Historical Hourly Wind Data and Corresponding Hourly Loads by Location a Net Load Exceedance Curve Can Be Constructed There Is a chance that all wind turbines produce zero power at the time of peak load There Is a chance that all wind turbines produce maximum power at the time of minimum load

The Firm Capacity Credit for Wind Can Be Based on a System Reliability Measure Exceedance Probability (%) 0100 Load (MW) Operating System Capacity Total Capacity + Peak Load Nuclear Cycling Coal Base Coal GT NGCC Firm Capacity Credit (% of Capacity) Wind: 5-20 Coal: Nuclear: NGCC: Without Wind With Wind Loss of Load Probability Reliability Increase with Wind Wind Firm Capacity Credit Engineering Guideline Typically 5% to 15% of wind turbine capacity is applied toward the reserve margin Years MW Capacity Credit Design Capacity Derated

Veselka GridSchool 2010 A Lot of Wind Capacity Is Needed to Meet Renewable Portfolio Standards (e.g., 20% Energy) Capacity (GW) EXISTING CAPACITY THERMAL CAPACITY TO BE ADDED A lot of wind capacity is needed to get a relatively small capacity credit WIND CAPACITY CREDIT (20%) In this example, wind installed capacity is greater than the thermal capacity additions

Veselka GridSchool 2010 In Addition to Hourly Operations, Variable Resource Technologies Will Affect both the Amount and Type of New Thermal Capacity Built in the Future Levelized Cost ($) Capacity Factor (%) GT NGCC Coal Nuclear Normalized Load (%) Nuclear Coal NGCC GT Exceedance Probability (%) Without Wind With Wind 100

Veselka GridSchool 2010 The “Optimal” Expansion Solution in Terms of Economics Can Be Approximated with More Sophisticated Mathematical Models Pre-planning –existing plus committed units Planning period (20+ years) –first year an uncommitted unit could operate Post-planning period –operate plants past last year –compute salvage value A Dynamic Programming (DP) Algorithm Is One Method for Solving Problems Time Years State (Expansion Option) “Best” Plan Over Time Important Considerations  Existing grid resources  Unit operating flexibility  Ancillary services  Wind variability & uncertainty  Technical minimum output levels  Transmission constraints  Load profiles and uncertainty  Fuel costs  …..