Summer School

Domaine de Normont, Dourdan - June 18-20 2018

Confirmed lecturers: René Aïd (Université Paris-Dauphine), Rüdiger Kiesel (Universität Duisburg-Essen), Mathilde Mougeot (Université Paris-Diderot & ENSIIE), Sean Meyn  (University of Florida)

Venue

The summer school is open to all students (Master’s degree or PhD) interested in the topics of the thematic semester. It will include:

  • A General Introduction to Electricity Markets (8h)- A Course by Pr. René Aïd (Université Paris-Dauphine) and Pr. Rüdiger Kiesel (Universität Duisburg-Essen)
Outline

  • Part 1 by Pr. Aïd: this talk will consist in two lectures. The first lecture will present the fundamentals of electric system to enable the audience to understand the specificities of the commodity "electricity" and the common features of electricity markets (forward, day-ahead, intraday and capacity markets). The second lecture will describe some problems that may have an interest for statisticians, namely (i) demand response programs (ii) retail market competition (iii) electricity usage identification.
  • Part 2 by Pr. Kiesel: Intraday Trading

    We will start with an introduction to the Day-Ahead and Intraday market as organised by the EPEX-Spot. After that an empirical analysis of the Intraday market will be presented and possible mathematical models will be discussed. In the final part we will present some results on optimal execution and market making.

Slides Pr. Aïd I

Slides Pr. Aïd II

Slides Pr. Kiesel 

  • "Reinventing Control and Economics in the Power Grid" (6h) - A Course by Pr. Sean Meyn (University of Florida)

Abstract

A firm belief throughout these lectures: economic efficiency means first understanding the nature of an efficient outcome. A description of this requires a definition of efficiency, and an understanding of dynamics of the grid as well as the electric loads that provide services to consumers. Does a person taking a shower care about the precise shape of the power trajectory that is heating the water? An understanding of the answer to this questions opens doors for better managing the grid. Common questions today: Greater communication and greater distributed resources: is this a blessing or curse? How expensive is volatility from renewables expensive? The answers depend on how you control the system.

Outline

  • 1. Motivation and Models: This lecture will focus on the distributed control architecture used in the power grid today, and the dynamic nature of electricity and markets. With respect to control, there are similarities with flight control: two remarkable examples of engineered reliability, which is the product of carefully designed feedback loops. A crash course on classical control will take up the first half of the lecture, with the second half devoted to the distributed control architecture in existence today (“AGC” + “Droop”). It will be clear that today's control architecture is sensible, but something is missing.
  • 2. Shock Absorbers for a Volatile Grid: In Lecture 1 we learned about many concerns of grid operators regarding power supply and demand, such as peaks, ramps, and forecast error. It is clear that an economic and reliable grid will require new approached to control, as well as new resources. Today there is interest in battery technology, and “demand response”. Most electric loads are flexible, and this flexibility can be harnessed to provide virtual energy storage. Techniques to quantify the capacity of this storage will be developed, based in part on mean-field models (a concept from statistical physics, but applied here in a very simple form).
  • 3. Demand Dispatch: The combined flexibility of power consumption in the European Union can smooth much of the electrical volatility from wind and solar generation. In this lecture we consider what control architecture can be put in place to make this possible with high reliability, and without excessive communication. As one example, we exploit the fact that the person in the shower cares about hot water and energy bills, and not about the precise shape of the power trajectory traveling through the wires behind her. A collection of one million hot water heaters can easily adapt its behavior to serve both the grid and the shower occupant. This adaptation requires new distributed control techniques.

Slides Pr. Meyn: - part 1 - // - part 2 - // - part 3 -

 

  • "Statistical and Machine Learning Methods to Model and Forecast Energy Consumption or Production" (6h) - A Course by Pr. Mathilde Mougeot (ENSIIE, LPSM)

Abstract

Since electricity can hardly be stored, forecasting tools are essential to appropriately balance consumption and production of energy, including renewable energies.   Analyzing historical data shows that time series of energy production or consumption may be radically different. Consequently, adapted statistical tools and methods should be used to model or forecast energy in both cases.

Outline

  • Part 1. The prediction box for energy consumption. This lesson introduces functional regression and sparse models and shows how these models can be used to forecast energy consumption. It will present the “prediction box” to forecast the French national energy consumption, a statistical model allowing to forecast in a high dimensional framework the intra day load curves of the French national consumption.
  • Part 2. Machine learning models to model wind energy production. On the other hand, to model and to forecast the wind energy, machine learning and aggregation techniques appeared to be more appropriate. This lesson presents several parametric and non parametric methods to model wind energy production, such as regression trees, bagging, random forest, boosting. Different methods for aggregation will be also introduced to take advantage of all models.
  • Part 3. Monitoring and diagnosis of overconsumption for industrial equipment. This lesson presents a methodology to monitor an industrial equipment based on regression models and scoring.

Slides: Introduction // Part 1 // Part 2 // Part 3

References

  • Cadet O., Harper C. and Mougeot M. (2005) Monitoring Energy  Performance of Compressors with an innovative auto-adaptive  approach. ISA 2005, Chicago.
  • Hastier T., Tibshirani R., Friedman J. (2009) The elements of statistical learning. Springer.
  • Fischer A., Montuelle L., Mougeot M., Picard D. (2017) Statistical learning for wind power: a modeling and stability study towards forecasting. Wind Energy.
  • Mougeot M., Picard D., Lefieux V., Maillard-Teyssier L. (2015) Forecasting intra day load curves using sparse functionnal regression. Springer Lecture Notes in Statistics, p. 161-182.
  • Mougeot M., Picard D., Tribouley K. (2013) Sparse approximation and fit of intraday load curves in a high dimensional framework. Advances in Adaptive Data Analysis, p. 1-23.

 

Tentative Programme

Monday 18th

9:00 - 10:00: Welcoming of the participants
10:00 – 12:00 A General Introduction to Electricity Markets (I), Pr. René Aïd
12:00 – 13:30 Lunch
13:30 – 15:30 A General Introduction to Electricity Markets (II), Pr. Rüdiger Kiesel
15:30 – 16:00 Coffee Break
16:00 – 18:00 A General Introduction to Electricity Markets (III), Pr. Rüdiger Kiesel
19:00 Buffet Dinner

Tuesday 19th

8:30 - 10:30 A General Introduction to Electricity Markets (IV), Pr. René Aïd
10:30 - 10:45 Coffee Break
10:45 - 12:45 Reinventing Control and Economics in the Power Grid (I), Pr. Sean Meyn
12:45 - 14:00 Lunch
14:00 - 16:00 Reinventing Control and Economics in the Power Grid (II), Pr. Sean Meyn
16:00 - 16:30 Coffee Break
16:30 - 18:30 Statistical and Machine Learning Methods to Model and Forecast Energy Consumption or Production (I), Pr. Mathilde Mougeot
19:30 Dinner

Wednesday 20th

8:30 - 10:30 Reinventing Control and Economics in the Power Grid (III), Pr. Sean Meyn
10:30 - 10:45 Coffee Break
10:45 - 12:45 Statistical and Machine Learning Methods to Model and Forecast Energy Consumption or Production (II), Pr. Mathilde Mougeot
12:45 - 14:00 Lunch
14:00 - 16:00 Statistical and Machine Learning Methods to Model and Forecast Energy Consumption or Production (III), Pr. Mathilde Mougeot
16:00 - 16:30 Coffee Break
16:30 - 19:30 Social Event
19:30 Dinner

Group picture I

Group picture II

Application

The students willing to apply can fill the online application form.

The (accepted) students will have all their travel and accommodation expenses covered by the organisation, for the summer school and for the closing conference that will directly follow the event.

If you have any question regarding this event, you can send an email to: damien[dot]fessler[at]dauphine[dot]fr