Key points are not available for this paper at this time.
Distributed optimization methods have been widely studied in recent years as they could be instrumental in optimizing the operation of future power systems. Compared to the centralized approach, communications will have greater impact on distributed methods as their performance depends heavily on the information exchange among multiple control entities. However, it is not clear what communications infrastructures and technologies should be adopted for distributed methods in practice. To investigate this issue and bridge the gap between the power system communications research community and the distributed optimization research community, this paper models the communications infrastructures and schemes required by distributed methods using the OPNET modeler. The capabilities and limitations of centralized and distributed communications infrastructures are investigated in terms of communications delays incurred and their impact on the convergence of two distributed algorithms, namely, Alternating Direction Method of Multipliers (ADMM) and Optimality Condition Decomposition (OCD) for solving the AC Optimal Power Flow problem in the IEEE 118-bus system and in the large-scale Polish power system. Our simulation results show that the existing communications plane in power systems might not be able to support widely used distributed optimization techniques (such as ADMM and OCD), which is a major finding. In addition, our results indicate that the connectivity requirements of distributed optimization necessitate a mesh network architecture with additional communications links among substations.
Guo et al. (Thu,) studied this question.