safewind advances in short-term wind power forecasting with focus on "extreme" ...

21
feWind ances in Short-Term Wind Power Forecasting with Foc "Extreme" Situations George Kariniotakis Head of Renewables & SmartGrids R&D Group MINES ParisTech - ARMINES [email protected] EWEA 2012,16-19 April 2012, Copenhagen, Denmark

Upload: vitalis-kiril

Post on 02-Jan-2016

26 views

Category:

Documents


0 download

DESCRIPTION

EWEA 2012,16-19 April 2012, Copenhagen, Denmark. SafeWind Advances in Short-Term Wind Power Forecasting with Focus on "Extreme" Situations. George Kariniotakis Head of Renewables & SmartGrids R&D Group MINES ParisTech - ARMINES [email protected]. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

SafeWindAdvances in Short-Term Wind Power Forecasting with Focuson "Extreme" SituationsGeorge KariniotakisHead of Renewables & SmartGrids R&D GroupMINES ParisTech - [email protected]

EWEA 2012,16-19 April 2012, Copenhagen, Denmark

Page 2: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

Introduction

2002

2010

2020

20 GW

74.7 GW

230 GW

2

Source: EWEA

• Ambitious targets for wind integration in EU

• Challenges for the power system management

• Short-term forecasting of wind generation is recognized as a prerequisite for an efficient wind integration

Page 3: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

0

2000

4000

6000

8000

10000

12000

00:0

001

:15

02:3

003

:45

05:0

006

:15

07:3

008

:45

10:0

011

:15

12:3

013

:45

15:0

016

:15

17:3

018

:45

20:0

021

:15

22:3

023

:45

time

Po

wer

[M

W]

measurementprediction

Introduction

• The actual wind power forecasting technology is quite mature

• However, in some situations large forecast errors may have an important impact on power system operation

Example in Germany:

Path of low-pressure system was different than predicted,

Maximum error: 5500 MW…

3

Page 4: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

Introduction

• The actual wind power forecasting technology is quite mature

• However, in some situations large forecast errors may have an important impact on power system operation

• The SafeWind project was developed to improve wind power predictability in challenging and extreme situations.

– at various temporal scales (very short to longer term)

– at various spatial scales (local/regional/continental)

4

Page 5: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

5

The SafeWind Project

9 countries22 partners

2008-2012

Coordination ARMINES/

MINES ParisTech

End-users

http://www.safewind.eu

Research

Industry

MeteorologistsINT

ER

NA

TIO

NA

L, I N

DIA

Page 6: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

6

Scientific & Technical Objectives (1)

• New forecasting methods for wind generation focusing on uncertainty and challenging situations/extremes:

o Probabilistic models

o Event forecasting models (ramps, cut-offs)

o Spatio-temporal correction models

o Regime switching models

o Prediction scenarios

o …

– Methodologies for communicating probabilistic forecasts & risks

Page 7: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

Highlight: Ramps forecasting using ensembles

New approach for probabilistic prediction of ramp events:• Prediction of ramps occurrence and timing. • Based on ensembles. • Ramps detection through a derivative filtering based on edge detection theory.

Results:• Controllable tradeoff between

ramp capture and forecast accuracy

• Reliable probabilistic forecasts with sharpness

• 5-15% performance increase with respect to climatology

7

Page 8: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

Highlight: Scenarios and their verification

Scenarios of wind power generation are a must-have for high-value decision-making, either• Using meteorological ensembles as input,

• Or based on probabilistic forecasts and spatio-temporal statistical models

Results :• A methodology for the

comparative verification of scenarios

• Methods to produce scenarios based on ECMWF ensemble forecasts or state-of-the-art probabilistic predictions

• Readily operational and used in practice

8

Page 9: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

Highlight: Spatio-temporal modelling for forecasting

New spatio-temporal forecasting approaches for the shorter term:• Wind farms are considered as measurement points• A spatio-temporal model is fitted on each wind farm using off site measurements• Improvement of wind power forecasts up to 18% (red when east, blue for west).

9

Case: Denmark

Page 10: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

10

Scientific & Technical Objectives (2)

• Develop a "European vision" for wind power forecasting– Prepare the way for the coordinated management of 200+ GW

wind generation at European Scale.

Data from Synoptic Stations in Europe

Monitor the Wind Energy Weather over Europe in real-time with observation data from many different sources

• More than 2000 weather stations• 120 single wind farms• 19 regions• 1 met mast

SafeWindData Management System

Page 11: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

11

Scientific & Technical Objectives (3)

• Models for "alarming" for very short-term (0-6 h). Monitor/assess the wind energy weather situation over Europe Detect severe deviations in forecasts due to extreme events Issue alerts to users that a forecast error is occurring Produce improved updates of the prediction in the short-term

Alarming in Forecast Time Series (23/1/2009)for Eastern part of Germany (50 Hertz)

Pressure gradient deviation (forecast - AdHoc-Analysis)

Page 12: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

12

Scientific & Technical Objectives (4)

• Models for "warning" on the level of predictability in the medium-term (next day(s)).

Approaches based on ensembles and weather pattern identification

Page 13: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

13

Scientific & Technical Objectives (5)

• Develop research in meteorology oriented to wind power forecasting: – Deliver the meteorological component for skilful

probabilistic wind power forecasts:• look at various products specifically suited for wind energy and

extremes (i.e. new product - 100m winds) (next talk)• evaluate various ensemble forecasting configurations• wind energy oriented verification of ECMWF products• improving ensemble forecasts (wind & wind power)

– Stimulate the interaction between meteorologists and wind power communities

Page 14: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

Highlight: Recalibrated ECMWF wind ensembles

Ensemble forecasts of wind speed and direction are of great value as input to wind power prediction• They need advanced post-processing (due to bias, lack of spread, etc.)

• They can be recalibrated with adaptive statistical techniques

Results:• ECMWF ensemble forecasts of

(u,v)-winds recalibrated for the whole Europe over the period 2007-2009

• Substantial improvement of various skill scores and diagnostics

14

Page 15: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

15

Scientific & Technical Objectives (6)

• Evaluate the impact of short-term wind predictability in the resource assessment phase:– when wind farms participate in an electricity market. – study if a tradeoff between wind potential and predictability can be

beneficial when selecting a site or expanding a portfolio

Terrain complexity

Flat Medium Complex

Wind potential

Predictability

A B

Page 16: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

1 & 2 day-ahead wind power forecast error for Central Europe• Computed from COSMO-EU Forecasts and Analysis (German Weather Service)• Normalization with estimated wind energy yield RMSE [MWh/MWh]• Spatial fc error smoothing considered (here: size of a DSO control zone)

Highlight: Forecast Error Maps

day-ahead day-ahead

16

2day-ahead 2day-ahead

See Posters PO 51 & PO109

Page 17: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

Highlight: Extreme Wind Atlases-from global data to site use

New approach to obtain turbine design winds • Downscale the global data to obtain extreme winds for turbine sites. • Use mesoscale modelling to simulate the yearly strongest storms over areas of

different terrain complexity.• Use a post-processing to prepare the data for microscale modelling.

Results :• Extreme wind atlases for places of

various extreme weather mechanisms and terrain complexities.

• Reasonable agreement with measurements.

• Output ready for microscale modelling for particular turbine sites.

Generalized extreme wind atlas for Navara region in Spain

17

m/s

Page 18: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

• Analyse how new measurement technologies like Lidars can be beneficial for better evaluation of external conditions, resource assessment and forecasting purposes.

Høvsøre Large Wind Turbine Test Facility

Scientific & Technical Objectives (7)

Measurement campaigns at flat (DK) and complex (ES) terrains

18

Page 19: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

Anemos.eXtreme: On-line demo of new approaches

• On-line demonstration of SafeWind modules:– Integration of 8 new modules into the ANEMOS wind

power prediction platform– Demonstration /evaluation for various TSOs

• Test cases:SONI (N. Ireland/U.K.) EirGrid (Ireland)

EDF (France) RTE (France)

PPC (Crete, Greece)

19

Page 20: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

20

Conclusions

• Successful scientific production: >75 papers so far

• New approaches for forecasting in challenging situations

• The work methodology designed to enable quick transfer of results for operational use by industrial stakeholders

• Final project workshop: 31 August 2012, Paris

What comes next :

• Improving predictability requires continuous collaborative R&D

• Need to better integrate forecasts into power system management tools

• R&D in forecasting is promoted as a priority (i.e. through TPWind).

Page 21: SafeWind Advances in Short-Term Wind Power Forecasting  with Focus on  "Extreme"  Situations

21

Thank you for your attention