A Rolling Horizon Method for a Bilevel Demand Side Management Problem
Résumé
With the emergence of the smart grid paradigm, combined with the breakthrough of distributed generation, demand side management has become more important than ever. Since distributed generation is by nature unpredictable, scattered, and irregular, the electricity distribution system undergoes many constraints. It it thus essential to be able to control the demand (and the rest of the production) as much as possible. This work focuses on one specific tool to achieve an efficient management of the demand: time-of-use pricing. Offering differing prices to customers who are going to optimally adapt their consumption is indeed a good way to induce load shifts that are beneficial: in this case, they aim to improve the revenue of an electricity furnisher.
The presented problem is based on Afsar et al.*, but incorporates the management of storage capacities and distributed generation, e.g. solar panels. Furthermore, scenario trees are considered to handle the unpredictability of the distributed generation. However, the inherent complexity of bilevel problems, combined with the potentially huge size of scenario tree-based models, makes the problem hard to solve. This talk addresses the problem with a rolling horizon method, whose efficiency is validated with numerical tests using real data.