Evaluation of Optimization Methods for Discrete Power Scheduling Applications in City Districts

Inhalt:
This paper evaluates three methods for the distributed optimization of power consumption and generation profiles of participants in city districts with discrete control decisions. The evaluated methods are based on the Alternating Direction Method of Multipliers algorithm. To yield the full operational flexibility in the distribution grid, discrete control decisions of devices on the demand side must be taken into account during their operation planning. An aggregation service provider which is involved in this planning needs to solve a Mixed-Integer Programming problem when utilizing flexibility of devices. This optimization can become computationally challenging. Scheduling algorithms that try to solve those problems centrally are typically not highly scalable. Instead, distributed methods can be used to create subproblems which can be solved in parallel on multiple computing nodes. All methods except for the Heuristic Release-and-Fix Method are actually intended to solve convex optimization problems. However, they are able to yield more accurate scheduling results for large city district scenarios compared to a state-of-the-art centralized method. Although the Heuristic Release-and-Fix Method was designed for solving loosely coupled Mixed-Integer Programming problems, the obtained schedules for the use-case scenarios in this study are considered worse than the schedules of the other distributed methods.