George Kariniotakis,

1
Towards Smart Integration of Wind Generation

• George Kariniotakis,
• Ecole des Mines de Paris/ARMINES, France
• georges.kariniotakis_at_ensmp.fr

2
Introduction
2002 2006 2010 2020
20 GW 48 GW 75 GW 180 GW (!)

• Reliable large-scale integration
• Economic and secure management of power systems
• Competitiveness of wind energy in a liberalised
  electricity market

Challenges
3
Context

• Wind power forecasting is recognised today as a
  necessary tool for 'smooth' wind integration
• Considerable research carried out in the last 15
  years
• "accuracy"-driven approaches.
• (i.e. models tuned based on least mean
  square error type of criteria).

Meteorology
Operational decision making
4
Context

• Wind power forecasting is recognised today as a
  necessary tool for 'smooth' wind integration
• Considerable research carried out in the last 15
  years
• "accuracy"-driven approaches.
• (i.e. models tuned based on least mean
  square error type of criteria).

Meteorology
Operational decision making
5
Context

• End-users today request also
• high reliability of prediction services and
  systems,

Meteorology
Operational decision making
6
Context

• End-users today request also
• high reliability of prediction services and
  systems,

• Challenge "value"-driven forecasting
  approaches.

Meteorology
Wind power forecasting technology
Operational decision making
7
The Project
2007-2010
8 countries, 23 partners
Budget 5.7 Mio
Coordinator Ecole des Mines de Paris, ARMINES,
France. Dr. George Kariniotakis,
http//www.anemos-plus.eu
8
Improvements in Wind Power Forecasting
Objectives

• Enhanced reliability, robustness ergonomy
• Multi-NWP approach
• Multi-models/combined approach
• Intelligent handling of situations with
  incomplete information
• Adaptability of prediction tools to changing
  environment
• (i.e. evolution of installed capacity)
• Extreme weather conditions

9
Improvements in Wind Power Forecasting
Objectives

• Extended functionalities
• Very short-term predictions
• Probabilistic forecasting (including ensembles)
• Advanced uncertainty estimation (i.e. for
  regional forecasting)
• Prediction risk indices for warning.
• Standardisation
• Interfaces, Data handling, Security aspects,
  Alarming,

10
Optimal Power System Management
Objectives

• Focus on functions like
• Reserves estimation
• Congestion management in large or local grids
• Power system scheduling
• Optimal coordination of storage and wind power.
• Optimal trading strategies based on probabilistic
  forecasts and prediction risk.

11
Optimal Power System Management
Objectives

• Power systems are traditionally operated using
  "determinstic" tools.
• It is necessary to adapt management functions for
  high wind penetration
• In practice, simplified approaches are used (
  adaptation of deterministic models).
• Some approaches consider wind variability
  estimates rather than wind predictability.
• Others try to model forecasting errors rather
  than using realistic timeseries or distributions.
  
• The aim is to develop operational tools based on
  the stochastic paradidm.

12
 
Objectives
Demonstration

• Demonstrate the performance of the advanced wind
  functionalities and resulting benefits.
• Demonstrate the benefits from the application of
  decision support tools based on stochastic
  analysis and optimisation. Comparison to actual
  practices.

13
Objectives
Demonstration
Conventional decision making/management
CO2
Dbenefits
Input
decision making/management
14
System Reserve Estimation
Example
Doherty, R. and OMalley, M.J., New approach to
quantify reserve demand in systems with
significant installed wind capacity, IEEE
Transactions on Power Systems, Vol. 20, pp. 587
-595, 2005.
15
Example
Optimal Trading
100
95
90
Revenue ( of max revenue)
85
80
75
70
Simple model
Perfect prediction
Probabilistic model Strategy 2
Probabilistic model Strategy 1
Persistence
Trading example Increase of benefits by the use
of advanced bidding strategies
P. Pinson, C. Chevallier, G. Kariniotakis,
"Optimal Strategies for Trading Wind Electricity
Markets using Probabilistic Wind Generation,
IEEE Transactions on Power Systems.
16
 
Objectives
Demonstrations
 
17
Conclusions

• Although in the last 15 years wind power
  forecasting was concentrating on deterministic
  approaches, today, state-of-the-art includes
• Advanced uncertainty estimation of deterministic
  models
• Ensembles approach
• Multi-model (NWP or WPP) approach and combination
• Prediction of uncertainty (risk indices)
• Probabilistic forecasting (p.d.f. prediction)
• more to come )
• Demonstration of the new approaches at
  operational conditions is needed (i.e. how to
  present all this information to end-users).
• It is necessary to bring to every day use
  state-of-the-art tools for managing efficiently
  power systems with high wind penetration (based
  on the stochastic paradigm) that integrate in a
  coherent way wind predictions and information on
  their uncertainty.

18

• Thank you for your attention

The ANEMOS.plus project is funded in part by the
European Commission under the 6th Framework
Program.