Systems Operations. PRAMOD JAIN, Ph.D. Consultant, USAID Power the Future. Astana, September, /6/2018

Transcription

1 Systems Operations PRAMOD JAIN, Ph.D. Consultant, USAID Power the Future Astana, September, /6/2018

2 Economics of Grid Integration of Variable Power FOOTER GOES HERE 2

3 Net Load = Load Wind Production Source: NREL FOOTER GOES HERE 3

4 Economic Dispatch What is the reserve requirement? 2x250 2x150 1x100 Source: P. Jain, Wind Energy Engineering, 2016 FOOTER GOES HERE 4

5 Impact of Variable Power on Conventional Generators in Unit Commitment 1. Higher grid inertia and governor response 2. Modest amount of higher system-wide reserves, depends on accuracy of RE forecast 3. Higher ramp rate of conventional generators 4. More frequent start, stop and lower dispatch levels of some conventional generator FOOTER GOES HERE 5

6 Need for higher grid inertia & governor response Source: P. Jain, Wind Energy Engineering, 2016 FOOTER GOES HERE 6

7 Need for higher grid inertia & governor response High wind penetration implies dispatch of lesser number of conventional generators Most wind* and solar power plants do not provide inertial or droop response Source: P. Jain, Wind Energy Engineering, 2016 FOOTER GOES HERE 7

8 Reserve Requirement for Different Forecast Errors Wind forecast is 100% accurate: No change in reserve requirement Wind forecast is 90% accurate No wind forecast is done Net load FOOTER GOES HERE 8

9 Ramp rate impact FOOTER GOES HERE 9

10 How to reduce reserves? Move from hourly to sub-hourly dispatch FOOTER GOES HERE 10

11 Illustrative Example: Generator Details Generator Capacity, Minimum Ramp rate, Type Marginal Cost, MW load, MW MW/min $/MWh A Base-load 50 B Base-load 55 C Flexible 58 D Flexible 58 E Flexible 60 Case 1, Wind farm Wind 0 Case 2, Wind farm Wind 0 FOOTER GOES HERE 11

12 Illustrative Example: Common data across all cases Maximum load ramp rate = 15 MW/min; Two scenarios of total installed wind capacity=100mw and 300MW Maximum wind ramp rate with 100MW of total wind install = 5 MW/min Maximum wind ramp rate with 300MW of total wind install = 15 MW/min Reserve requirement (RR) = 100MW + 10% of total installed wind capacity High wind scenario FOOTER GOES HERE 12

13 Illustrative Example: Assumptions for this simplified Unit Commitment Start-up and shut-down cost Fixed cost Start-up and shut-down time Minimum run- and down-time. FOOTER GOES HERE 13

14 Illustrative Example: Dispatch for 3 scenarios No wind RR=100MW Installed wind capacity=100mw, WF = 100%, RR=110MW Installed wind capacity=300mw, WF=100%, RR=130MW 6AM, Load=800MW A,B,C A,C,D B,C,D 9AM, Load=1,000MW A,B,C,D A,B,C,D A,C,D 11AM, Load=1,100MW A,B,C,D A,B,C,D A,C,D FOOTER GOES HERE 14

15 Illustrative Example No wind RR=100MW Installed wind capacity=100mw, ED WF = 100% RR=110MW Installed wind capacity=300mw, ED WF=100% RR=130MW 6AM, Load=800MW A=500,B=270,C=30 A=500,C=175,D=25 B=300,C=175, D=25 9AM, Load=1,000MW A=500,B=300,C=175, D=25 A=500,B=300,C=75, D=25 A=500,C=175, D=25 11AM, Load=1,100MW A=500,B=300,C=200, D=100 A=500,B=300,C=175, D=25 A=500,B=270, C=30 FOOTER GOES HERE 15

16 Grid Flexibility Before new flexible generators are procured, the hidden flexibility in the existing network should be fully exploited Generators are profitable when dispatched at close to 100% Generators are profitable when there is very little cycling Generators do not want to incur shut down and start up cost Minimum loading requirement of IPP generators (Contractual) Demand: Dispatchability of load to exploit flexibility in load Increasing balancing area: Pooling reserves Shorter scheduling intervals FOOTER GOES HERE 16

17 Conclusions Changes to System Operations is the least expensive solution to wind and solar PV grid integration issues 17

18 AGENDA What is VRE forecasting? Definitions Process of VRE forecasting Properties of VRE forecasting Why is it needed? Wind forecasting methods FOOTER GOES HERE 18

19 What is VRE forecasting? VRE forecasting is the short-term (dealing in hours and days) prediction of the amount of power that will be generated by the VRE power plant in the future. Short-term generally refers to week-ahead (WA), day-ahead (DA) or intraday (ID). Examples of ID are hour-ahead (HA), 15 minutes-ahead (15MA) and 5 minutes-ahead (5MA). The duration of ID may be few hours or until end of the day. 19

20 Definitions of VRE Forecasting Terms Day-ahead forecast Hour-ahead forecast 20

21 Definitions RMSE = σ t=1 n (f t a t ) 2 n MAD = σ n t=1 f t a t n MAPE = MAD C 100 where, t is a time block, f t is the forecast for time block t, a t is the actual observation for time block t, and n is the number of time blocks for which the error is computed. C is the available capacity of the VRE facility 21

22 Process of VRE forecasting Weather Forecast Historical Generation & Weather Data VRE Forecasting Algorithm VRE Generation forecast Availability Forecast 22

23 Role of VRE forecasting in dispatch planning Load Fx VRE Fx Available Cap Fx Demand Response Total Load Total Available Generation Schedule for all generators, exchange/ sale/ purchase of power Input Dispatch Planning Output 23

24 Properties of VRE forecasting The accuracy of the VRE forecast increases as the forecast time block and lead time decrease. It should be noted that only a grid with fast dispatching can capitalize on the higher accuracy The accuracy of the VRE forecast increases as the number of VRE plants is increased and as the geographical diversity is increased The amount of total regulation required depends on VRE forecast accuracy; greater accuracy leads to less regulation The accuracy of the VRE forecast depends primarily on the accuracy of both the weather forecast and available capacity data 24

25 Ecosystem of VRE forecasting Centralized forecasting Done by System Operations VRE forecasts at pooling substation Best practice Decentralized forecasting Done by VRE plant Effective if penalties exist for high deviations Hybrid approach Done by both 25

26 Why is VRE Forecasting Important for Grid Integration of VRE? Flexibility is a system s ability to respond to the variability and uncertainty of supply and demand. 26

27 Impact of fast schedule and dispatching on reserve requirements 27

28 Wind Forecasting Methods Physical approach Based on accurate wind flow around and inside the wind farm, in addition to using the manufacturer's power curve in order to propose an estimation of the wind power output. It consists of several sub models, which together deliver the translation from the NWP forecast at a certain grid point and model level, to power forecast at the considered site and at turbine hub height Statistical approach Emulates the relation between meteorological predictions, historical measurements and generation output through statistical models whose parameters have to be estimated from data, without taking any physical phenomena into account. Hybrid approach Combine the two approaches in order to join the advantages of both approaches and thus improve the forecasts Source: Dr Balaraman, Wind Forecasting presentation 28

29 Physical approach The two main steps are downscaling and conversion to power Source: Dr Balaraman, Wind Forecasting presentation 29

30 Statistical approach Source: Dr Balaraman, Wind Forecasting presentation 30

31 Statistical approach Persistence Models: Persistence forecasting assumes that the power at a certain future time will be the same as it is when the forecast is made The very short-term forecasting approach consists of statistical models based on the time series approach, such as the Kalman Filters, Auto-Regressive Moving Average (ARMA), Auto- Regressive with Exogenous Input (ARX), and Box-Jenkins forecasting methods. For time horizons greater than 6 hours, NWPs would be used as inputs 31

32 Approaches to aggregate forecasting Aggregate forecasts for a region or pooling substation are often done by System Operator Inputs: Availability forecast from individual VRE plants NWP weather predictions for a few representative locations in region Historical production data from the region or pooling substation Aggregation of wind farms reduces the forecast error as a result of spatial smoothing effects Methods in Regional Forecasting; Direct Upscaling Cascaded Approach Cluster or Sub regions Approach 32

33 USAID Regional Program Power the Future Pramod Jain, Consultant President, Innovative Wind Energy, Inc. Thank You Power the Future 6, Sar y Arka Ave, Office 1430 Astana, Kazakhstan DISCLAIMER This product is made possible by the support of the American People through the United States Agency for International Development (USAID). The contents of this presentation are the sole responsibility of Tetra Tech ES, Inc. and do not necessarily reflect the views of USAID or the United States Government. 33