OPERATIONS and DISPATCH GE Energy Asia Development Bank Wind Energy Grid Integration Workshop: OPERATIONS and DISPATCH Nicholas W. Miller GE Energy Consulting Beijing September 22-23, 2013 This lecture is closely related to the economics discussion: operation and economics are interrelated. In the end, the system must be dispatched both economically and reliably. Disclaimer: The views expressed in this document are those of the author, and do not necessarily reflect the views and policies of the Asian Development Bank (ADB), its Board of Directors, or the governments they represent. ADB does not guarantee the accuracy of the data included in this document, and accept no responsibility for any consequence of their use. By making any designation or reference to a particular territory or geographical area, or by using the term “country” in this document, ADB does not intend to make any judgments as to the legal or other status of any territory or area.

ADB topic list Wind energy dispatching methodology ( International Expert) • Wind farm as Capacity source or Energy source • Policies for scheduling wind energy • Policies for curtailing wind energy • Comparison different methods • Software and other tools, processes for scheduling wind energy • Role of Wind energy forecasting

Time Scales for System Planning and Operation Processes The system needs to be run, i.e. dispatched, to cover all the time scales within a few days…. The system needs to be PLANNED to have the right equipment in place to allow successful – economic and reliable – dispatch.

System Operation Process - Overview Day Ahead Prepare load forecast (Total MW load for each hour of the day) Commit units that will run to serve the load (accounts for uncertainty) Preliminary dispatch schedule for each unit (by hour) Units with long startup times are “committed” for operation during the next day Hour Ahead Perform hour-ahead load forecast Adjust hourly dispatch for committed units as required to match actual load Real Time Load-following (typically, dispatch is adjusted at 5-minute intervals) Adjustments based on “economic dispatch”, using marginal costs or competitive bids Regulation (fast adjustments of MW to regulate frequency and intertie power flows)

For grid operations, wind is “similar” to load . Like load, wind can be forecast a day ahead Grid operators can plan day-ahead operations base on a load forecast and a wind generation forecast Dispatchable generation is allocated to serve the net of the forecast load minus the forecast wind Uncertainty in the wind forecast adds to the uncertainty in the load forecast Adjustments are made using hour-ahead forecasts and real-time data Dispatchable Generation Serves “Net Load” Net Load = Load Minus Wind (This is what must be served by other types of generation) Departure from “net load” approach is usually very expensive.

Overview Temporal/Spatial Patterns Variability in Wind and Load MW Uncertainty Forecasting for Wind Power

Monthly Energy GWh from Wind & Solar for Years 2004–2006 (30% Wind Energy - In Area Scenario) ‘06 ‘05 ‘04 Notable difference in Wind & Solar energy across the months and over the years

55% of energy from wind and solar Monthly Energy % from Wind & Solar for Years 2004–2006 (30% Wind Energy - In Area Scenario) 06 55% of energy from wind and solar ‘06 30% is not always 30% ‘05 ‘04 Policy people think “% energy” on an annual basis. Dispatch people think % power on an instantaneous basis There is huge difference in perspective. 2006 percent monthly energy ranges from 18% (July) to 55% (April) in study footprint

Study Footprint Total Load, Wind and Solar Variation Over Month of July (30% Wind Energy in Footprint) LP Scenario Wind may or may not exhibit daily cycles. Under high load conditions with low wind, load dominates

Study Footprint Total Load, Wind and Solar Variation Over Month of April (30% Wind Energy in Footprint) LP Scenario But when it is very windy, like in the spring (april) And, the load is low, the wind variability dominates This is what operators worry about

Study Footprint 2006 Net Load Duration – In Area Scenario 0 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Deciles of Year Min load 22169 MW Below existing min load ~57% of year, for 30% scenario EVERY system worries first about light load, high wind…. The minimum load condition is the most difficult Being about to shutdown and dispatch down baseload generation becomes ABSOLUTELY critical

What Does 30% Penetration Mean? At the 30% annual energy, some hours might approach 100%

Overview Temporal/Spatial Patterns Variability in Wind and Load MW Uncertainty Forecasting for Wind Power

Variability Analysis - Deltas Statistics used to characterize variability: Delta (∆) – The difference between successive data points in a series, or period-to-period ramp rate. Positive delta is a rise or up-ramp Negative delta is a drop or down-ramp Mean () – The average of the deltas (typically zero within a diurnal cycle) Sigma (σ) – The standard deviation of the deltas; measures spread about the mean For a normal distribution of deltas, σ is related to the percentage of deltas within a certain distance of the mean

Average Daily Profile of Deltas Over Year 2006 (30% Wind Energy in Footprint – LP Scenario) (Avg. +/- sigma, Minimum, Maximum) 5644 MW (Nov 14) Load and Net Load Delta (MW) Total Load and Net Load (MW) The net load variability from wind increases But not uniformly…. In this system, more reserves are needed especially around hour 1600. Economic operation requires good data… do not add reserves uniformly across all hours. -4931 MW (Jun 7) Hour of Day

Overview Temporal/Spatial Patterns Variability in Wind and Load MW Uncertainty Forecasting for Wind Power

Standard Deviations of Day-Ahead Forecast Errors 33,000 MW Peak Annual Load 3,300 MW Total Wind Plant Rating 950 MW With Wind 800 MW Without Wind Load forecasts AND wind forecasts are always wrong. It is the total error that is a problem for dispatch

Overview Temporal/Spatial Patterns Variability in Wind and Load MW Uncertainty Forecasting for Wind Power

Forecasting Wind forecasting is absolutely essential Forecasting increases economic value of wind power by >25% or more Wide-spread extreme wind events are predictable (e.g. widely publicized Texas events were predicted) Texas February 24, 2007 event Arrival of such fronts is generally forecastable, several hours ahead within a 30-minute window Extreme Thirty-Minute Wind Drops ~1600 MW Multiple system studies (NY, Ontario, Colorado, Minnesota, Idaho, California, Alberta,…) have all reached the conclusion that maintaining and developing a portfolio that includes substantial (and increasing) levels of flexible generation is critical to enabling high levels of wind penetration Systems with limited hydro resources will need to rely on the latest technology thermal plants to meet these requirements. Existing hydro resources need to maintain their flexibility against encroachment (bank erosion, fish kill, irrigation, etc.) The value of flexible generation to resource owners (and grid operators) is rapidly becoming as critical as heat rate and other more traditional measures of value; flexibility is a competitive advantage Each MW of flexible generation will enable successful integration of many MWs of additional wind generation (exact level is system dependent) ~1.5hours

Reserve Requirements

Large System Net Load Variability : Separating Wind and Load Effects (30% case) 10-minute Ds Net load variability increases with wind Implied reserve requirement is 3 x Ds Requirement is a function of both load level and wind level Wind Level Distillation of 3Ds to Simple Rule: X% of Load plus Y% of Wind Production with a max The industry is still experimenting with different ways to specify additional reserves. This is one example from recent research. The plot looks complex, but the end result can be simple rules for dispatchers to add reserves based on the wind forecast. Load Level

Can’t lean on the Neighbors Wind Power Data: Extrema More Important in Small Systems Provided by AWS Truewind (2 years of 10-min for each plant) 500 MW Wind case (inc 2 x 200MW remote island plants Can’t lean on the Neighbors Small systems that have little or no connection to neighbors need to be more conservative with their reserves New Reserve = Spin + Up Regulation = 185MW + f (Wind)

Flexible Generation

Dealing with Variability Balance of generation portfolio (dispatchable generation) must have the capability to respond to variations in net load Net load = (Load MW) – (Wind MW) Generators must have room to maneuver up or down Ramp RANGE up and down Generators must be capable to maneuver fast enough to follow changes in net load Ramp RATE (MW/minute) The following slides show how Ramp Range and Ramp Rate for an operating area are affected by increasing penetration of wind generation

Grid maneuverability decreases as wind penetration increases 10% Wind Energy 20% Wind Energy Week of April 10, Spring Season Load levels are typically low Wind generation is typically higher in spring than other seasons Wind plant output is typically greater at night Grid has difficulty operating at “minimum load” 30% Wind Energy This is a detail of the minimum load problem. Wind plants can be selectively curtailed to eliminate this problem. Dispatchers MUST have the right and capability to occasionally curtail wind plants. There MUST be a penalty assigned to system operation, NOT just the wind plants, for curtailment. Market mechanisms are one way to do this.

Subhourly Time Simulations QSS (Quasi Steady-State) Simulations vs. LTDS (Long-term Dynamic Simulations) Provide Validation and Context for Operational and Statistical Analysis Cases Selected from Statistical Analysis Boundary Conditions Set by Operational Analysis Evaluate Impact of Significant Wind Generation Load Following & Ramp Rate Requirements Regulation/AGC Requirements Illustrate Performance Issues Illustrate Mitigation Measures

Unit-Type Dispatch 20% Wind max min 30% Wind 10% Wind AS combine cycle plants are pushed out, coal plants (steam) must cycle more. Ramp rate limits on plants can become very important.

Dealing with Uncertainty Basic options are increased reserves or demand response Increasing reserves Commit additional generation so that load will never be interrupted Need to do it 100% of the time, because you do not know the reserves will be required Demand response Interrupt or reduce load occasionally, as need arises A paid ancillary service

Using Load to Meet Occasional Extremes Load Interruption This shows how occasionally – about 1% of all hours – curtailing a small amount of load (about 1-2%) can avoid the need for ver conservative, and expensive oeration strategies. Load Energy PAID to be interupted

Costs are per MWh of energy reduced. Interruptible loads are easily cost justified The cost benefit of occasionally interupting loads can be huge in high wind systems. In this example, on the right hand side, curtailing 1 MWH is “worth” $100,000. Cost of reducing Unserved Energy by discounting wind generation forecasts. (i.e., adding reserves in proportion to forecasted wind generation) Costs are per MWh of energy reduced.

Impact on the Existing Generation Fleet? Lower capacity factors for base and mid-merit generation Use of “peakers” at “unusual” times Pressure to increase hydro maneuverability Increased combined cycle cycling (today and growing rapidly) Increased coal cycling (growing rapidly in some places) Increased O&M, higher outage rates, environmental performance impacts Credible quantitative data is limited; sensitive Claims of costs, loss of life, and physical capability are variable Severity of impacts and the allocation of costs is a topic of intense debate

Capacity Value of Wind Generation

Effective Capacity or Effective Load Carrying Capability (ELCC) ELCC is a measure of long-term adequacy Ability of a plant to serve load Avoid loss of load by the power grid Example of a 100 MW thermal plant If forced outage rate is 10%, and If forced outages are equally probable at any time, then ELCC is 90% How does this measure apply to wind power? Output of a wind plant is not dispatchable Wind plant output is a function of available wind, and it is time-dependent Capacity value is best calculated with Monte Carlo simulations of impact on unserved load. This method is widely used for conventional generation in the US. The method works well with wind generation, too. 33 Source: WindLogics

2001 Average Load versus Average Wind August July September

Effective Capacity Based on rigorous LOLP calculations using 2001 - 2003 load and wind profiles for NY State Inland Wind Sites: Capacity factors ~ 30% Effective capacity, UCAP ~ 10% Offshore Wind Site: Capacity factors ~ 40% Effective capacity, UCAP ~ 39% Developed approximate calculation method: UCAP ~ On-Peak Capacity Factor for 1:00-5:00pm, June-August 684 358 570 322 400 261 105 600

Experience and Lessons Learned GE Energy Experience and Lessons Learned

Major Study Results : Get the infrastructure right And use it better Large interconnected power systems can accommodate variable generation (Wind + Solar) penetration levels exceeding 30% of peak loads But not by doing more of the same….. To reach higher levels of wind generation and other renewables: Get the infrastructure right And use it better The debate has changed: No longer: “Is it possible?” Now: “How do we get there?”

Missing Wind/Solar Target Higher Cost of Electricity Lessons Learned Impediments Lack of transmission Lack of control area cooperation Market rules / contracts constraints Unobservable DG – “behind the fence” Inflexible operation strategies during light load & high risk periods System Cost Unserved Energy Missing Wind/Solar Target Higher Cost of Electricity Renewables (%) Enablers System Cost Impediments Enablers Wind Forecasting Flexible Thermal fleet Faster quick starts Deeper turn-down Faster ramps More spatial diversity of wind/solar Grid-friendly wind and solar Demand response ancillary services All grid can accommodate substantial levels of wind and solar power … There is never a hard limit