On the effects of communication topologies on the performance of distributed optimization heuristics in smart grids

Abstract

Distributed heuristics have shown promising results in handling difficult optimization and coordination tasks in smart grids, which often have to deal with large numbers of components, distributed information and real time constraints. Some of these heuristics are distributed on an algorithmic level. This means that the way information is exchanged between the distributed units can effect the efficiency of the algorithm, in terms of computational effort, message volume, and convergence speed. It may even impact the effectiveness, i.e. the solution quality. The performance of control systems is of utmost importance for stable and optimal operation of critical infrastructure in smart grids. Therefore, factors influencing the effectiveness and efficiency, like the communication topology, must be thoroughly investigated in order to both initially configure them optimally and react appropriately to undesired behavior at runtime. The impact of the communication topology is studied with an experimental setup, using COHDA as an example heuristic. Systematic experiments are performed with various topologies and varying numbers of agents to illustrate the importance of a solid and maybe even dynamic choice of the communication topology for distributed heuristics in smart grid applications.