2011.01.21 Wind Workshop Notes (courtesy of Karl Critz) Charles Smith Executive Director Utility Wind Integration Group; http://www.uwig.org/ Reston, VA markets - low cost/low flexibility flexible generation traditional storage wind curtailment (transmission tradeoff. tx 17%!) storage - high cost/high flexibility time scales: regulation (seconds-minutes) load following (minutes-hours) unit commitment (hours-days) ISOs - not much difference day-ahead, hour-ahead, 5-min big diff within-hour 5-minute range. (caused largely by variability in ancillary services models) Idea: differentials in LMP between midwest and east could pay for transmission upgrades. (BUT it could just act as an export mechanism for coal power.) balancing cost - compare wind penetration and $/MWh forecasting for day-ahead operations situational awareness forecast - severe weather hour-ahead: 10 min forecasts 4-6 hours ahead (updated hourly) (reliability) day-ahead forecast: hourly forecasts 2-4 days ahead, updated 12 hours (economic) forecast error, single plant energy (% actual) 10-15% capacity (% rated) 4-6 day ahead 24-30% hourly day ahead capacity grr ERCOT Feb 26 2008 event. actual generation exceeds 80% bounds on low side fossil plant failed to come online (800MW) unexpected load increase (1000MW) aggregation - reduces variability. Wind is an energy resource, not a capacity resource. (Need capacity to meet reliability.) value of storage - remains roughly constant as penetration increases. induction drive machines give damping for frequency response. google transmission line - personally not necessary to build DC cable. Suspects it's better to connect directly with the AC grid. Michael Brower, CTO, AWS Truewind awstruepower.com; wind energy forecasting - most isos get forecast data $4.50/MWh integration cost at 10%. forecasts can reduce uncertainty, reduce cost plant power curves nonlinear, so small wind errors = big power error 5-60 minutes - regulation/realtime dispatch 1-6 hours - load-following 24 hours - unit commitment, scheduling, market trading seasonal/long-term - resource planning, contingency analysis measuring forecasts - mean error, mean absolute error, rms error mean absolute error 1 hr 6-7% 2 9% 3 12-15% 4 12-16% day ahead MAE eWind ~15% ensemble forecast - weighted mean of multiple forecast regime forecasts - "characteristic regimes", optimized for each regime ramp forecasting - where to measure - may need to measure 300km away! need hig-altitude measurements too. integration - - confidence levels grid operators don't care about small errors, they care about big ramps that could bring the grid down. forecast errors larger when forecasting a ramp Michael Milligan - NREL - cost of integration for wind and solar energy: large -scale studies and implications; WWSIS, EWITS "large" integration - ~10%. (have studied to 30%, rough ideas at 40%) tradeoffs: build near wind or build near people. wind power vs transmission. varies by month - 30% annual penetration varies between 15 and 55% sysop does "net load" load - renewables (treat renewable as exogenous and optimize around full implementation) requires coordinated unit commitment and dispatch value with perfect forecasts vs state of the art (see study) See RISOE for stochastic model, interesting: EWITS puts no wind in the south. variability over large geographic distances - aggregation impact incredible. most variability at the middle of nameplate. At high end, on the flat part of the curve. At the bottom end, not much change. "If you love wind power, you have to at least like transmission" (like Neko Case) every MWh of wind offsets something. Need to be prepared to be nimble with thermal generating units. for integration - less baseload more flexible storage varies ERCOT gets 50% of load response from demand response. Bill Henson - ISO New England - NEWIS - 4GW transmission overlay, 10GW wind short-term forecast error drives majority of required regulation requirements (more than wind variability) using more wind means that regulation resources are used more often. offshore wind tends to correlate more with peak load in NE Mark O'Malley , Ireland - Republic of Ireland & Northern Ireland unified power system 9.7 GW total, 1.8 GW wind max load 6.5GW, min load 2.4 GW, 450VDC link to britain highest wind penetration of any synchronous power system more than denmark (E/W denmark are separate and part of larger systems) In ireland, wind and load are pretty synchronous. nice. (plot: average load over the year by hour) unfortunately, highest peak ever recorded was on a near-zero wind day. capacity factor 1999-2009 31% (max 34%, min 29%) 2010 - 23% (historical low) when ireland adds a lot of wind, it reduces more co2 in GB than ireland! (stops imports from british coal, starts exporting irish gas) stochastic unit commitment. "Unit commitment for Systems with Significant …" o'malley et al good paper to read. (downloaded) how do you embed the cost of starts and cycling (long-term costs) into a short-term optimization? (most interesting research question) storage - 1 pumped hydro unit (292 MW for 5 hours) Out on maintenance for 9 mo no additional storage in the initial integration study. [pumped hydro is 70-85% round-trip efficient] studied up to 12GW (80%) with stochastic optimization, storage leads to less curtailment. assertion: storage device will always cost more than a thermal unit (turbine+civil eng!) with more storage, in ireland it will use more british coal to charge its storage units (Eric Ingersoll asserts that having 300 hours of storage enables seasonal shifting, not just day-to-day shifting.) doubly-fed induction generator wind turbine. a bit elastic. doesn't add any inertia to the system. harder to do frequency regulation. the more wind there is on the system, the more frequency dips you get with an event GE has a mechanism in its new turbines to help with this. ---- ge slides ---- nicholas.miller@ge.com, GE Wind, INERTIA - use controls to extract stored rotational inertial energy, don't slow the blade too much, or stall watch multiple control loop interactions note, however, that the turbine will use power to regain its kinetic energy; wind turbines can theoretically be better for frequency control than thermal (due to greater rotational inertia) grid codes - must be able to respond to sustained undervoltage Eric Ingersoll, CEO, general compression long-duration storage for baseload wind goal: displace nuke and coal. there is no market for storage tech. biz model: sell power we are recruiting raised $24M oh no: hard to compete vs fully-depreciated coal plant developers don't get capacity credit for wind policy supports weak and undependable ITC - (investment tax credit) - 30-40% of capex (?) good: capex coming down turbines getting better v112: 99m blade! microforecasting (10sec lookahead) O&M coming down goal: compete in "firm" power market utilities don't want to buy wind helps transmission - greater capacity factor issues: duration of storage, variation in resource , no storage: to make a 100MW contract, build 200MW. 8-10hr storage: miss 1800hr/yr dispatch, dump 18% of energy (insufficient storage) 200 hr storage: no dumping, no hours missed usually need 100-300hr storage, depending. more capital-intensive than gas plants. near-zero marginal cost to run tech: compresser/expander, use salt caverns quasi-isothermal compression system. little thermal loss. $48M, can build salt cavern stores 3000-4000 MWh $4/kWh TX plant earns $35/MWh intermittent, $70/MWh firm (TX nuke $80/MWh) irony: ireland & new england bring in wind, backs off clean gas. requires long discharge time, high system power. compresser-expander cost relates to power cavern cost is discharge time not terribly related .