7th session of seminars on research in energy economics at paris-sciences-lettres

Date

13 novembre 2013

Lieu

University Paris Dauphine-PSL

Description

MODELLING DYNAMIC OPTIMIZATION OF INVESTMENT IN COMPETITIVE ELECTRICITY MARKETS UNDER UNCERTAINTY

Wednesday November 13, 2013, from 16h30 to 18h30
University Paris-Dauphine, Place du Maréchal de Lattre de Tassigny,
Room A 2-A 3, 2th floor

PROGRAMME AND SLIDES

The Seminar on Research in Energy Economics at Paris-Sciences-Lettres is jointly organized by the CERNA, the CGEMP, the Chaire European Electricity Markets, Mines ParisTech and University Paris-Dauphine. It is animated by François LEVEQUE (CERNA et MINES PARIS TECH) and Dominique FINON (Chaire European Electricity Markets, CNRS-CIRED).

Michaela UNTEUTSCH (Energy Research Institute EWI, Cologne University), Optimization of power plant investments under uncertain renewable energy development paths – A multistage stochastic programming approach
Presentation
Working paper

Andreas EHRENMANN, (Senior Analyst, GDFSuez Direction of Strategy, Brussels), Stochastic Equilibrium Models for Generation Capacity Expansion
Presentation
Andreas EHRENMANN and Yves SMEERS’Paper

Summary of the presentations

Michaela UNTEUTSCH
Electricity generation from renewable energy sources (RES-E) is supposed to increase significantly within the coming decades. However, uncertainty about the progress of necessary infrastructure investments, public acceptance and cost developments of renewable energies renders the achievement of political plans uncertain. Implementation risks of renewable energy targets are challenging for investment planning, because different RES-E shares fundamentally change the optimal mix of dispatchable power plants. Specifically, uncertain future RES-E deployment paths induce uncertainty about the steepness of the residual load duration curve and the hourly residual load structure. In this paper, we show how uncertain future RES-E penetrations impact the electricity system and try to quantify effects for the Central European power market. We use a multi-stage stochastic investment and dispatch model to analyze effects on investment choices, electricity generation and system costs. Our main findings include that the uncertain achievement of RES-E targets significantly affects optimal investment decisions. First, a higher share of technologies with a medium capital/operating cost ratio is cost-efficient. Second, the value of storage units in systems with high RES-E penetrations might decrease. Third, in the case of the Central European power market, costs induced by the implementation risk of renewable energies seem to be rather small compared to total system costs.

Andreas EHRENMANN
Capacity expansion models in the power sector were among the first applications of operations research to the industry. The models lost some of their appeal at the inception of restructuring even though they still offer a lot of possibilities and are in many respect irreplaceable provided they are adapted to the new environment. We introduce stochastic equilibrium versions of these models that we believe provide a relevant context for looking at the current very risky market where the power industry invests and operates. We then take up different questions raised by the new environment. Some are due to developments of the industry like demand side management: an optimization framework has difficulties accommodating them but the more general equilibrium paradigm offers additional possibilities. We then look at the insertion of risk related investment practices that developed with the new environment and may not be easy to accommodate in an optimization context. Specifically we consider the use of plant specific discount rates that we derive by including stochastic tic discount rates in the equilibrium model. Linear discount factors only price systematic risk. We therefore complete the discussion by inserting different risk functions (for different agents) in order to account for additional unpriced idiosyncratic risk in investments. These different models can be cast in a single mathematical representation but they do not have the same mathematical properties. We illustrate the impact of these phenomena on a small but realistic example.