Power Systems & Energy Course Large-Scale Wind and Solar Integration: Dealing with Variability and Uncertainty Jason MacDowell An Introduction to Today’s Lectures… Focus on utility-scale wind and solar What’s different about wind and solar? Mitigating operational impacts Key lessons learned from studies and operational experiences 2016International, General Electric All Rights © 2016 General © Electric Inc AllCompany rights reserved Not forReserved distribution without permission Successful integration of high penetrations of wind IMPORTANCE OF BUILDING ON SUCCESSES IN OTHER REGIONS AND LEARNING FROM THEIR MISTAKES Source: DOE/LBNL, 2015 Wind Technologies Market Report © 2016 General Electric International, Inc All rights reserved Not for distribution without permission Moderate annual average penetration means high instantaneous penetrations Source: AWEA (2/22/16) and Ventyx (top) S Beuning, Xcel, 2011 (bottom) © 2016 General Electric International, Inc All rights reserved Not for distribution without permission © 2016 General Electric Company All Rights Reserved We can successfully integrate high penetrations of solar Source: Rothleder, CAISO, UVIG 2016 © 2016 General Electric International, Inc All rights reserved Not for distribution without permission © 2016 General Electric Company All Rights Reserved How you effectively integrate wind and solar? Cost-effective, efficient integration of wind and solar Can be done through markets (RTOs) or verticallyintegrated utilities Actions that increase operational efficiency are generally the same actions that help integrate renewables Systems approach is critical Requires cooperation: utilities, RTOs, developers/owners, regulators/policymakers © 2016 General Electric International, Inc All rights reserved Not for distribution without permission © 2016 General Electric Company All Rights Reserved What Makes Wind/Solar Different? Zero marginal cost • Cycling impacts on other generation • Cost recovery impacts on other generation System Balancing • Variability and uncertainty • Load is also variable and uncertain Reliability and stability • Inverter-based, non-synchronous generation • Essential reliability services • Weak grid Location • Remote, with long transmission • Distributed energy resources, behind the meter © 2016 General Electric Company All Rights Reserved © 2016 General Electric International, Inc All rights reserved Not for distribution without permission Overview • Temporal/Spatial Patterns • Variability in Wind and Load MW Solar Variability • Uncertainty • Forecasting for Wind Power Value of Improved Forecasts in Grid Operations © 2016 General Electric International, Inc All rights reserved Not for distribution without permission GE’s Integration of Renewables Experience 2010 New England Studies commissioned by utilities, commissions, ISOs • Examine feasibility of 100+ GW of new renewables • Consider operability, costs, emissions, transmission 12 GW Wind 39% Peak Load 24% Energy 2008 Maui 70 MW Wind 39% Peak Load 25% Energy 2010 Oahu 500 MW Wind 100 MW Solar 55% Peak Load 25% Energy 2012 NSPI Study 900MW Wind 25% Energy PJM Study (underway) 96GW Wind 22GW Solar 30% Energy Gradients indicate systems subject to individual studies and also included in larger regional studies 2004 New York GW Wind 10% Peak Load 4% Energy 2005 Ontario 15 GW Wind 50% Peak Load 30% Energy 2006 California 13 GW Wind GW Solar 26% Peak Load 15% Energy 2007 Texas 15 GW Wind 25% Peak Load 17% Energy 2009 Western U.S 72 GW Wind 15 GW Solar 50% Peak Load 27% Energy Pan-Canadian ~72GW Wind 30% Energy Universal need for fleet flexibility, new operating strategies and markets, transmission reinforcement, grid friendly renewables © 2016 General Electric International, Inc All rights reserved Not for distribution without permission Planning and Operation Process Year Unit Dispatch 700 Resource and Capacity Planning (Reliability) Capacity Valuation (UCAP, ICAP) and Long-Term Load Growth Forecasting 600 500 MW Slower (Years) Time Scales for System Planning and Operation Processes Technology Issues 400 300 200 100 0 2000 4000 6000 8000 Hour 2001 Average Load vs Average Wind Day 30,000 1,600 1,400 1,000 15,000 800 600 10,000 400 Wind Output (MW) 1,200 20,000 5,000 July load August load July w ind 200 Septem ber load August w ind Septem ber w ind 0 11 16 21 Hour 3000 Hours 2500 Load Following (5 Minute Dispatch) Faster (seconds) Day-ahead and Multi-Day Forecasting NYISO Load (MW) Unit Commitment and Day-Ahead Scheduling Hour-Ahead Forecasting and Plant Active Power Maneuvering and Management 2000 MW Time Frame 25,000 1500 1000 500 61 121 M inu te s September Morning Frequency and Tie-Line Regulation (AGC) Real-Time and Autonomous Protection and Control Functions (AGC, LVRT, PSS, Governor, V-Reg, etc.) © 2016 General Electric International, Inc All rights reserved Not for distribution without permission A ugus t Morning May Ev ening Oc tober Ev ening April Af ternoon 10 Minutes 10 Overview • Temporal/Spatial Patterns • Variability in Wind and Load MW Solar Variability • Uncertainty • Forecasting for Wind Power Value of Improved Forecasts in Grid Operations © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 44 Variability and Uncertainty Variability: Wind and solar generator output varies with the wind and sun • Several timescales - minute, hour, diurnal, seasonal Uncertainty: Wind and solar generation are similar to load • Not dispatchable – output is predicted by a forecast • Actual power output is different from forecast output A perfect forecast eliminates uncertainty, but there is still variability 25000 2500 20000 2000 15000 1500 CSP Actual MW MW CSP Forecasted 10000 1000 Wind Actual 5000 500 Forecasted Wind 0 12 18 24 Hours 30 36 42 48 12 18 24 Hours © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 30 36 42 48 45 Forecast Error: Load Only and Load+Wind 3000 Load Alone Load Error (F-A) MW Error 2000 2000 MW 1000 -1000 49 97 145 193 241 289 Hour 3000 Load + Wind Total Error (F - A) MW Error 2000 2500 MW 1000 -1000 49 97 145 Hour 193 241 289 33,000 MW Peak Annual Load 3,300 MW Total Wind Plant Rating © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 46 Overview • Temporal/Spatial Patterns • Variability in Wind and Load MW Solar Variability • Uncertainty • Forecasting for Wind Power Value of Improved Forecasts in Grid Operations © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 47 Operators’ Tools Source: Wayne Moodie, PJM, UVIG Feb 2014 Source: Signoretty, ALSTOM, UVIG Feb 2013 Source: Pengwei Du, ERCOT, UVIG Feb 2015 © 2016 General Electric International, All rightsAllreserved Not for distribution without permission © 2016 General ElectricInc Company Rights Reserved 48 Wind Power Forecasting for Power System Operation Source: © 2016 General Electric International, Inc All rights reserved Not for distribution without permission Atmospheric Complexity The atmosphere is so complex… So how does this work? Solar Radiation Convection Moisture Fluxes Condensation Turbulence Evaporation 50 Surface Heat © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 50 Gridded 3D Weather Data Integrates all available data sources, from the surface to the upper atmosphere, into a unified and physically consistent state of all grid cells at a given point in time Over 160 weather variables collected from: • Surface / METAR station data • Oceanographic buoys • Ship reports • Aircraft (over 14,000 ACARS/day) • NOAA 405 MHz profilers • Boundary-layer (915 MHz) profilers • Rawinsondes (balloon soundings) • Reconnaissance dropwinsonde • RASS virtual temperatures • SSM/I precipitable water • GPS total precipitable water • GOES precipitable water • GOES cloud-top pressure • GOES high-density vis cloud drift wind • GOES IR cloud drift winds • GOES cloud drift winds • VAD winds: WSR-88D NEXRAD radars 51 © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 51 Meteorological Models Numerical gridded representation of the laws of physics • Conservation relations – – – – Mass Energy Momentum Water, etc • Physical processes – Radiation – Turbulence – Soil/ocean interactions, etc • Use lots of fast computers – Partial differential equations – Gridpoint difference values – Step all points through time using very small steps (a few seconds per step) 52 © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 52 Forecasting Technology & Results Advanced physical and statistical methods • Multiple forecast models & computational learning systems • Adaptive adjustment based on local observations Data management is critical • Getting the wind plant data can be the challenge Results for a typical wind plant • Next day hourly power: 10-14% MAE of rated capacity • Next day total energy: 20% MAE of energy delivered • Next 2-3 hour power schedule: 5-7% MAE of rated capacity When aggregated on a system-wide basis, errors are substantially reduced • Often 30-50% depending on geographic dispersion • “Portfolio effects” become important & help operations 53 © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 53 Source: WindLogics 54 Confidence Intervals & Decision Support 55 © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 55 Overview • Temporal/Spatial Patterns • Variability in Wind and Load MW Solar Variability • Uncertainty • Forecasting for Wind Power Value of Improved Forecasts in Grid Operations © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 56 Forecasting saves money Forecasting saves up to 14% in annual operating costs If forecasts were perfect, an additional 1-2% could be saved © 2016 General Electric International, Inc All rights reserved Not for distribution without permission Source: GE Energy, Western Wind and Solar Integration Study 2010 57 Model and methodology improvements reduce forecast errors Ana Rodriguez, REE, UVIG, 2015 © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 58 [...]... Wind Increases 4000 Summer Wind Delta (MW) (30% Scenario) 3000 (-3339,1611) 2000 Δ = 3674 Maximum for Load Alone 100 0(-4250,135) 0 -100 0 (3674, -106 1) Δ = 4735 -2000 (2496,-2 210) Δ = 4706 -3000 -4000 Load Increases Wind Decreases Load and Wind deltas offset -5000 -5000 -4000 -3000 -2000 -100 0 0 100 0 Load Delta (MW) 2000 3000 © 2016 General Electric International, Inc All rights reserved Not for distribution... Generator Owner… “I can guarantee 100 0MW of hydro all day tomorrow.” System Operator… “OK, I will turn off 100 0MW of other generation Variability: Generator Owner “I can guarantee 100 0MW of hydro from 2PM to 4PM tomorrow.” System Operator… “OK, I may turn down 100 0MW of other generation, rather then shutting it off.” Uncertainty: Generator Owner… “I think I will have 100 0MW of hydro sometime tomorrow.”... Pattern: July 2003 Average Day (California) Average Solar 100 00 100 0 0 0 1 5 9 13 17 21 Hour © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 11 Temporal Pattern: January 2002 Average Day (California) 50000 5000 Average Load Average L-W-S Average Wind 4000 Average Solar 30000 3000 20000 2000 100 00 100 0 0 Wind & Solar (MW) Load (MW) 40000 0 1 5 9 13 17... and potential mitigation options for these impacts.” Wind and Solar Combinations (% Energy) Baseline: Existing Wind and Solar Generation 10% In-Area: 10% Wind, 1% Solar In Footprint 10% Wind, 1% Solar Out of Footprint 20% In-Area: 20% Wind, 3% Solar In Footprint 10% Wind, 1% Solar Out of Footprint 30% In-Area: 30% Wind, 5% Solar In Footprint 20% Wind, 3% Solar Out of Footprint Solar Mix: • 70% Concentrating... 30000 Min load 22169 MW 20000 Below existing min load ~57% of year, for 30% scenario 100 00 0 -100 00 00 10% 583 20% 1166 30% 1749 40% 50% 60% 2332 2915 3498 Deciles of Year Hour of Year 70% 4081 80% 4664 © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 90% 5247 100 % 5830 21 Operational Impact of Wind & Solar: What does 30% Penetration Mean? Nova... and wind deltas offset Load Decreases Wind Increases 4000 Summer Wind Delta (MW) (30% Scenario) 3000 2000 100 0 (-4250,-203) 0 (3674,-44) -100 0 Load Increases Wind Decreases -2000 -3000 -4000 MOST-CHALLENGING REGION FOR GRID OPERATORS Load and Wind deltas offset -5000 -5000 -4000 -3000 -2000 -100 0 0 100 0 Load Delta (MW) 2000 3000 © 2016 General Electric International, Inc All rights reserved Not for distribution... Days of July 2003 55000 15000 100 00 Load (MW) 40000 Significant day-to-day variation 25000 5000 100 00 0 1 5 9 13 Hour 17 Wind & Solar (MW) Average Load Average Wind Average Solar 21 © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 13 Temporal Pattern: All Days of January 2003 50000 100 00 8000 30000 6000 20000 4000 100 00 2000 0 Wind & Solar (MW)... Distribution of Extreme Footprint Hourly Net Load Deltas 2006, Local Priority Scenario 160 140 SA Baseline SA L-W-S (10% ) SA L-W-S (20%) SA L-W-S (30%) Number of 1-Hr Deltas 120 Curtail wind 100 To increase reserves 80 Curtail load To reduce need for additional resources For 30% penetration, 108 up-ramps of 3400 MW/hr or more 60 For 30% penetration, 5 down-ramps of 4200 MW/hr or more 40 For baseline, 3... Code 900 Zip Code 921 Zip Code 956 Zip Code 94a Zip Code 945 50 0 5 10 15 20 Concentrating Solar (CSP) is less variable because of thermal mass © 2016 General Electric International, Inc All rights reserved Not for distribution without permission 34 Island of Oahu (State of Hawaii) 1100 MW Peak Load 600 MW Minimum Load Planning to add: 100 MW Wind 360 MW Solar PV Potential Sites for Wind and Solar Generation... Daily Profile of Deltas Over Year 2006 (10% Wind Energy in Footprint – LP Scenario) (Avg +/- sigma, Minimum, Maximum) Load Deltas Net Load Deltas Total Load Total Net Load 8000 8000 40000 Average daily load profile 35000 4178 MW (Max 1-hr rise) 4000 4000 30000 Mean + Sigma Average L-W-S profile 2000 2000 25000 00 20000 -2000 -2000 15000 Mean - Sigma -4000 -4000 100 00 4195 MW (Max 1-hr drop) -6000 -6000 ... Alone 100 0(-4250,135) -100 0 (3674, -106 1) Δ = 4735 -2000 (2496,-2 210) Δ = 4706 -3000 -4000 Load Increases Wind Decreases Load and Wind deltas offset -5000 -5000 -4000 -3000 -2000 -100 0 100 0 Load... L-W-S (10% ) 50000 Study_Area L-W-S (20%) Study_Area L-W-S (30%) Net Load Level (MW) 40000 30000 Min load 22169 MW 20000 Below existing load ~57% of year, for 30% scenario 100 00 -100 00 00 10% 583... 100 0 (-4250,-203) (3674,-44) -100 0 Load Increases Wind Decreases -2000 -3000 -4000 MOST-CHALLENGING REGION FOR GRID OPERATORS Load and Wind deltas offset -5000 -5000 -4000 -3000 -2000 -100 0 100 0