PNNL

Abstract

The current practices for restoring critical services in the distribution grid during a disaster align with the traditional centralized ideology of distribution operations. A central processor evaluates the distribution grid after a disruption and sketches a restoration plan, which requires complete situational awareness. However, such a centralized operational paradigm is prone to single-point failure and computationally complex. This motivates a distributed decision-making paradigm where multiple agents solve smaller sub-problems and jointly coordinate their individual decisions to achieve the global/network-level objective. This paper aims to enable a distributed decision-making framework to support the bottom-up restoration of the distribution grid using all available resources, including distributed generation (DGs), while only requiring regional awareness and limited communications with neighboring connected regions. Towards this goal, we propose a Layered Architecture for Distributed Algorithms for Resilience (LADAR) for distribution system restoration. The proposed two-stage layered method is robust to singlepoint failures, enables autonomy using distributed algorithms, and reduces the computational burden of solving centralized optimization for a large-scale unbalanced distribution system.