Microgrid sizing and energy management using Benders decomposition algorithm

Microgrids are of increasing interest because they can facilitate the integration of renewable energy sources. To make the most of microgrids, optimization problems are formulated and solved to determine their optimal planning (i.e. sizing and energy management). However, these problems are complex and time-consuming to solve. In this article, we focus on a temporal decomposition based on Benders’ algorithm to reduce computing time while still obtaining the optimal solution. The temporal decomposition divides the initial problem into subproblems with a smaller time interval. The first originality of this work is the proposition of a methodology to apply this temporal decomposition to mixed-integer linear problems for the optimal planning of microgrids. The second originality is the investigation of the influence of the following relevant parameters on the computing time of the temporal decomposition based on Benders’ algorithm: decomposition period, nature of the problem, overall time horizon and number of CPUs. In addition, contrary to previous literature, our proposed method exhibits computing time reductions. They are of up to 5.6 times for the considered case studies. Our results also highlight the existence of a decomposition period that maximizes the performances. Besides, we find that the temporal decomposition is particularly efficient for mixed-integer linear problems with large time horizons and when more than 16 CPUs can be used. The proposed generic methodology and our results can notably be useful to researchers and to microgrids project holders who aim at finding the optimal sizing and operation of their microgrid within reduced computing time. •A mixed-integer linear problem (MILP) is solved to optimize microgrid planning.•Benders’ algorithm performs the temporal decomposition of the optimization problem.•Method applied to a case study where computing times are reduced up to 5.6 times.•There is a decomposition period that minimizes computing times.•Distributing evenly subproblems across CPUs is key for achieving high performance.