Demand-response aggregators are faced with the challenge of how to best manage numerous and heterogeneous Distributed Energy Resources (DERs). This paper proposes a decentralized methodology for optimal coordination of DERs. The proposed approach is based on Dantzig-Wolfe decomposition and column generation, thus allowing to integrate any type of resource whose operation can be formulated within a mixed-integer linear program. We show that the proposed framework offers the same performance guarantees as a centralized formulation, with the added benefits of distributed computation. The practical efficiency of the algorithm is demonstrated through extensive computational experiments, on a set of 1120 instances generated using data from Ontario energy markets. The proposed approach was able to solve all test instances to proven optimality, while achieving significant speed-ups over a centralized formulation solved by state-of-the-art optimization software.
Technical Report, Polytechnique Montreal, Department of Mathematics and Industrial Engineering, Montreal, QC, Canada & Groupe d’études et de recherche en analyse des décisions (GERAD), Montréal, QC, Canada