Domaine de Normont, Dourdan - June 21-22 2018
Confirmed speakers: Pierre Pinson (Technical University of Denmark), Tony Ware (University of Calgary), Delphine Lautier (Université Paris-Dauphine), Bertrand Villeneuve (Université Paris-Dauphine), Carlo Sgarra (Politecnico di Milano), Benoît Sévi (Université de Nantes), Nicolas Legrand (INRA), Alvaro Veiga (PuC, Rio), Jia-li Mei (Université Paris-Sud), Laurent Dubus (EDF Lab), David Salant (TSE), Philippe Quirion (CIRED), Laurent Lamy (CIRED)...
9:00-10:00 Welcoming of the participants
10:00-12:00 Session I - Auctions in Electricity Markets // David Salant (TSE), Laurent Lamy (CIRED) or Philippe Quirion (CIRED), Aurélie Nasse (EDF)
Laurent Lamy (CIRED) Estimating the incentive effect of auction-determined contracts
Abstract: Measuring how contractual characteristics induce contractors to make efforts and then to perform well is a difficult task because the determinants of the contracts are endogenous and the matching of agents to principals is not random. We develop a novel approach that is relevant for environments where contracts are the outcome of an auction process, and more specifically when the bids submitted by the principals for each given agent determine both the remuneration rules and the winning principal: it consists in modelling sources of unobserved heterogeneity in the performance equation in relation with private signals that enter an auction model that endogenizes the contractual form.
The econometric methodology proceeds then in two steps: The first step consists in estimating the primitives of an auction model (after having shown that the model is identified non-parametrically) which further allows to derive the distribution of the unobserved heterogeneity conditional on the bidding history. The second step consists in plugging the estimated expected values of the sources of unobserved heterogeneity as controls in the performance equation. In our most general specification, we consider a model with two kinds of unobserved heterogeneity: idiosyncratic synergies between agents and principals, and also a common value term that affects the performance of the given agent irrespective of the identity of the principal he is matched with. From the perspective of the auction literature, our model departs from the standard independent private value model (IPV) in two aspects: first, bidders' payoff functions are no longer linear functions of the auction price insofar as bidders are assumed to internalize that there is an incentive effect; second, we allow a given bidder to have information that matters for his competitors in the auction so that the model is one with interdependent values.
As an illustration, we will discuss the case of the contracts for renewable energies that are assigned through auction schemes and where we wish to measure possible detrimental incentives associated to existing remuneration rules which insure bidders against meteorological risks. This could provide an incentive to realize projects of smaller size that the quantity announced in the auction or equivalently to strategically misreport the expected quantity at the bidding stage.
David Salant (TSE), Auctions in Electricity Markets: An Overview
Abstract: I present a few of the main insights from auction theory and explain how they apply to electricity markets. The Revenue Equivalence Theorem (RET) indicates that the outcome, including revenues (or procurement costs) should, on an expected value basis, be the same in a pay as bid and uniform price auctions. Vickrey auctions have been recently used in the telecoms sector as they have the desirable feature that bidders will often have an incentive to bid true values. Unfortunately, Vickrey auctions have some undesirable properties as well.
Supply function bids are common in energy auctions. I explain how such auctions often have only mixed strategy equilibria. Finally, based on experience in sequential auctions for serving default service load, how the division of the auction volume across time can affect the outcome.
Aurélie Nasse (EDF) Auctions in Renewable Energy
Abstract: In order to decrease the subsidies for renewable energies, most countries have been moving from fixed tariff to competitive systems like auctions. In this presentation, we will go back quickly on the overall context and market of wind and solar, show the impact of auctions on the price, and with a couple of examples how the specific rules of auctions can have an impact on the offered price.
13:30-15:30 Session II - Smart Grids and Forecasting // Pierre Pinson (Technical University of Denmark), Maxime Grangereau (Ecole Polytechnique), Peter Tankov (ENSAE)
Pierre Pinson (Technical University of Denmark) High-Dimensional, Adaptive and Distributed Learning for Renewable Energy Forecasting
Abstract: Renewable energy forecasts with lead times up to a few hours are important for system operators and utilities to maintain a balanced and reliable power system. Commonly, these forecasts are computed by time series models that are mainly based on past data from the forecast site. More advanced models additionally employ data from surrounding sites or can adapt to changes in the weather regime or wind farm setup. The wealth of data generated by an increasing number of wind power installations does not only provide possibilities for improvements but also challenges for common forecasting methodologies. We will introduce, apply and discuss some of the recent proposals for high-dimensiona
Maxime Grangereau (Ecole Polytechnique) Stochastic Optimal Control of a Battery in a Micro-grid : Resolution using McKean-FBSDE
Abstract: We study the problem of the optimal control of a battery in order to reduce the costs associated to the management of the electricity network for the Distribution System Operator. Unlike previous works related to micro-grid management, we consider a system connected to the network and aim at limiting both the power peaks and the fluctuations of the power supplied by the network. This results in a mean-field stochastic optimal control problem, in a context of general filtration. We derive necessary and su cient optimality conditions, using the stochastic Pontryagin principle. This gives rise to a particular Mc-Kean-FBSDE. An existence result is obtained under technical assumptions for small time horizons. A particular focus is given to the Linear Quadratic case, where we can solve the optimality system for arbitrary time horizons and derive a closed-loop feedback formula for the optimal control. This formula involves the solutions of two backward Riccati Ordinary Differential Equations and two affine-linear Backward Stochastic Differential Equations. Joint work with Emmanuel Gobet.
Peter Tankov (ENSAE) Optimal Management of a Wind Power Plant with Storage Capacity
Abstract: We consider the problem of a wind producer who has access to the spot and intraday electricity markets and has the possibility of partially storing the produced energy using a battery storage facility. The aim of the producer is to maximize the expected gain of selling in the market the energy produced during a 24-hour period. We propose and calibrate statistical models for the power production and the intraday electricity price, and compute the optimal strategy of the producer via dynamic programming.
15:30-16:00 Coffee Break
16:00–17:20 Session III - Other Methods for Electricity Markets I // Joseph Mikael (EDF R&D), Emmanuel Gobet (Ecole Polytechnique)
Joseph Mikael (EDF R&D) Artificial Intelligence Applications in Energy Risk Management
Abstract: Traditional stochastic approaches for pricing and hedging have limitations (curse of dimensionality, restricted kind of underlying’s models…) that AI-based techniques promise to lift. We show on various example how neural networks performs regarding hedging problems. Moreover, we show how Neural Networks may help building prices model or realistic load curves.
Emmanuel Gobet (Ecole Polytechnique) Day-ahead Probabilistic Forecast of Solar Irradiance: a Stochastic Differential Equation Approach
Abstract: We derive a probabilistic forecast of the solar irradiance during a day at a given location, using a stochastic differential equation (SDE for short) model. We propose a procedure that transforms a deterministic forecast into a probabilistic forecast: the input parameters of the SDE model are the Arome deterministic forecast computed at day D-1 for the day D. The model also accounts for the maximal irradiance from the clear sky model. The SDE model is mean-reverting towards the deterministic forecast and the instantaneous amplitude of the noise depends on the clear sky index, so that the fluctuations vanish as the index is close to 0 (cloudy) or 1 (sunny), as observed in practice. Our tests show a good adequacy of the confidence intervals of the model with the measurement. This is a joint work with Jordi Badosa (LMD, Ecole Polytechnique), Maxime Grangereau (EDF and Ecole Polytechnique), and Daeyoung Kim (Ecole Polytechnique).
19:00 Conference Dinner
9:00-10:20 Session IV - Market Fundamentals: Equilibrium Modelling // Delphine Lautier (Université Paris-Dauphine) or Bertrand Villeneuve (Université Paris-Dauphine), Nicolas Legrand (INRA)
Delphine Lautier (Université Paris-Dauphine) & Bertrand Villeneuve (Université Paris-Dauphine) The Joint Dynamics of Spot and Futures Prices in Commodity Markets
Abstract: We give new insights into the dynamic behavior of commodity prices with an infinite hori- zon rational expectations equilibrium model for spot and futures commodity prices. Numerical simulations of the model emphasize the heterogeneity that exists in the behavior of commodity prices by showing the link between the physical characteristics of a market and some stylized facts of commodity futures prices.
Nicolas Legrand (INRA) The Delaying Effect of Storage on Investment: Evidence from The Crude Oil Sector
Abstract: Our paper provides a theoretical framework able to represent with accuracy a consistent relationship between fixed capital investment, storage and the term structure of prices in a storable commodity market. It aims at understanding the interaction of storage capacity with irreversible investment decisions in mediating investment and commodity price dynamics. The results show that the presence of storage, while smoothing the spot price tends also to channel volatility into the future, thereby raising the options value of waiting and eventually delaying and making lumpier the investment in fixed capital. The time-varying expected price volatility related to the inventory levels is a new channel we identify to show why irreversible investment decisions in a storable commodity market capture more accurately both price and investment dynamics observed in the data as compared to an irreversible investment setting without storage capacity. Joint work with Assia Elgouacem.
10:20 – 10:40 Coffee Break
10:40 – 12:00 Session V - Energy Price Modelling // Carlo Sgarra (Politecnico di Milano), Tony Ware (University of Calgary)
Carlo Sgarra (Politecnico di Milano) A Forward Price Model for Power Markets Based on Branching Processes
Abstract: We propose and investigate a market model for forward prices in power markets, including most basic features exhibited by previous models and taking into account self-exciting properties. The model proposed extends Hawkes-type models by introducing a two-fold integral representation property. A Random Field approach was already exploited by Barndorff-Nielsen, Benth and Veraart who adopted an Ambit Field framework for describing the forward dynamics. The novelty contained in our approach consists in combining the basic features of both Branching Processes and Random Fields in order to get a realistic and parsimonious model setting. We discuss the no-arbitrage issue of the present modelling framework. We outline a possible methodology for parameters estimation. We illustrate by graphical representation the main achievements of this approach. Joint work with Y. Jiao, C. Ma and S. Scotti.
Tony Ware (University of Calgary) Polynomial Processes for Modelling Energy Commodity Prices
Abstract: Energy commodities are notorious for the extreme dynamics exhibited by prices in spot and futures markets. Models for such prices typically include mean reversion, seasonality, as well as jumps and regime switching components. Polynomial processes have the property that conditional expectations of polynomial functions of future values are themselves given by polynomial functions of the current value. Examples include exponential Lévy processes, affine processes, and Pearson diffusions. In this talk we will explore how polynomial maps of polynomial processes can produce models that are able to represent the rich dynamics seen in energy commodity prices, while simultaneously providing easily-computable formulae for futures prices.
13:15-15:15 Session VI - Other Methods for Electricity Markets II // Jiali Mei (Université Paris-Sud, EDF R&D), Laurent Dubus (EDF R&D), Alvaro Veiga (PUC-Rio)
Jiali Mei (Université Paris-Sud, EDF R&D) Nonnegative matrix factorization with side information for time series recovery and prediction
Abstract: Motivated by the recovery and prediction of electricity consumption time series, we extend Nonnegative Matrix Factorization to take into account external features as side information. We consider general linear measurement settings, and propose a framework which models non-linear relationships between external features and the response variable. We extend previous theoretical results to obtain a sufficient condition on the identifiability of NMF with side information. Based on the classical Hierarchical Alternating Least Squares (HALS) algorithm, we propose a new algorithm (HALSX, or Hierarchical Alternating Least Squares with eXogeneous variables) which estimates NMF in this setting. The algorithm is validated on both simulated and real electricity consumption datasets as well as a recommendation system dataset, to show its performance in matrix recovery and prediction for new rows and columns.
Laurent Dubus (EDF R&D) Meteorological information to support energy decisions
Abstract: The energy sector has been one of the most important users of meteorological information in the last decades. New energy systems organization, the energy transition and the increasing share of renewable energy sources require improved datasets and forecasts. The talk will briefly describe what kind of data meteorology can provide, how these are obtained, and what can be expected from an energy perspective. We will then present current developments and plans for the next 5-10 years, and particularly the emerging field of climate services for the energy sector.
Alvaro Veiga (PUC - Rio) Vine Copulas for Simulating Inflows
Abstract: In the Brazilian Electrical Sector, energy is mostly generated by hydroelectric power plants. This causes an uncertainty in the operationalization of the system, making its planning non-trivial. As decisions are made under a high level of uncertainty, models based on stochastic programming are employed for this task. These models determine operational costs, energy prices, CO2 emissions, and system expansion. Usually, inflow scenarios feed these models, which represent all the uncertainty associated with energy production, and directly impact the decision‐making process. Thus, the realism of decisions made by the operator is dependent on the quality of the scenarios. The accuracy of an inflow simulation model can be determined by its capability to reproduce scenarios while preserving the main, historically recorded features, such as asymmetry or periodicity. Many inflow simulation models are built on rigid assumptions. This may limit their ability to represent nonlinear dependencies and/or nonstandard distributions. Copulas surmount these drawbacks and are a flexible tool for modeling multivariate distributions. They can separately model the marginal behavior of variables from the dependence structure of a random vector. Moreover, they can represent any type of relationship. The aim of our research is to present and discuss recent methodological developments, and present perspectives regarding the use of copulas for inflow-scenario simulation. Most of our work is based on Vine Copulas (Czado, 2009), used to deal with moderate to high dimensionality. Joint work with Guilherme A. A. Pereira.
15:15-15:30 Coffee Break
15:30-16:50 Session VII - Market Fundamentals: Empirical Research // Benedikt Gleich (Universität Augsburg), Benoît Sévi (Université de Nantes)
Benoît Sévi (Université de Nantes) Informed Trading in the WTI Oil Futures Market
Abstract: The weekly release of the U.S. inventory level by the DOE-EIA is known as the market mover in the U.S. oil futures market. We uncover suspicious trading patterns in the WTI futures markets in days when the inventory level is released that are higher than market forecasts: there are significantly more orders initiated by buyers in the two hours preceding the official release of the inventory level. We also show a drop in the average price of -0.25% ahead of the news release. This finding is consistent with informed trading. We also provide evidence of an asymmetric response of the oil price to oil-inventory news, and highlight an over-reaction that is partly compensated in the hours following the announcement. Joint work with Olivier Rousse.
Benedikt Gleich (Universität Augsburg) Fundamentals and Long Term Trends in Energy and Mineral Commodity Prices
Abstract: The recent mineral commodity boom has shown steep price increases, but has also been followed by calming markets. In comparison, prices for crude oil are on the rise again. By addressing systematic errors of classic real price analysis we provide a long term analysis of the "real real" price of mineral commodities and crude oil. We complement these results with additional empirical insights on the search for fundamental price drivers of commodities. Thus, we contribute to answer the question if oil is really special from an empirical point of view and if the current price fluctuations are historically common or indeed a sign of structural change.
16:50 End of the Conference
- Further information coming soon -