Title: Adaptive probabilistic forecasting of temporal data for electricity markets
Speaker: Yannig Goude – Senior Researcher: Machine Learning & Forecasting
- 2 – 3.30pm Seminar + Q&A in Room 2.41, Fry Building
- 3.30 – 4pm networking with students and colleagues in Common Room, Fry Building.
Location: Room 2.41, Fry Building
Optimising production and making decisions on the electricity markets requires the modelling of various hazards (electricity consumption, temperature, solar radiation, wind, hydraulic contributions, unavailability of power plants, market prices, and renewable production) based on historical data. For them to work and be of practical interest, in most cases these models have to meet a number of requirements: they have to be interpretable (understandable by the end users, generally engineers in the business departments), adaptive (in the sense of automatically adapting to changes in the data) and efficient. Against a backdrop of increasing production from renewable sources and major tensions in the markets (rising electricity prices, unavailability of production facilities, gas supplies), the teams at EDF R&D (Osiris department) are developing forecasting models that combine these different characteristics.
Yannig Goude is a senior researcher working at EDF R&D, in the department Optimisation Simulation RIsk and Statistics (Osiris), since 2008 and an associate professor at the Laboratoire de Mathématiques d’Orsay, Université Paris-Saclay. He obtained a Ph.D. in statistics and probability in 2007 at the Université Paris-Sud 11, Orsay. His research interests are forecasting, time series, machine learning, semi-parametric models, and online aggregation of experts.