Latest Improvements of EDF Mid-term Power Generation Management

Abstract

To optimize mid-term power generation management, Electricité de France (EDF) has developed a set of computer tools, which provide an order of magnitude of the supply and demand balance for the few following years. They also compute hydraulic reservoir management strategies for short-term unit commitment to correctly handle hydraulic reservoirs. This set of tools must be used every day with 500 scenarios and must give results in less than half an hour. Calculation durations are really critical. That is the reason why we had to study in depth the modeling and the optimization methods to reduce calculation durations as much as possible.In this set of tools, a new simulator has been developed with numerous improvements. Those improvements consist in a more precise generation unit modeling and a better generation unit commitment optimization. Those allow, for instance, to take into account piecewise linear Bellman function to optimize hydraulic reservoir commitment and to use either exact methods of resolution or several heuristics to optimize generation unit commitment. The first kind of heuristics begins with a choice of thermal power stations that are in use at a specific time, then allocates a part of the total production to each power plant. The second type is based on a branch and bound algorithm to solve a mixed integer program (MIP). Moreover, this article compares the performances of several linear solvers on an industrial problem.