Smart Load and Generation Scheduling for Power System Restoration

This paper studies the applicability of the linearized DC model in optimizing power restoration after significant network disruptions. In such circumstances, no AC base-point solution exists and the objective is to maximize the served load. The paper demonstrates that the accuracy of the linearized DC model degrades with the size of the disaster and that it can significantly underestimate active and apparent power. To remedy these limitations, the paper proposes an AngleConstrained DC Power Flow (ACDCPF) model that enforces constraints on the line phase angles and has the ability to shed load and generation across the network. Experimental results on N-3 contingencies in the IEEE30 network and power restoration instances from disaster recovery show that the ACDCPF model provides significantly more accurate approximations of active and apparent power. In the restoration context, the ACDCPF model is shown to be much more reliable and produces significant reduction in the size of the blackouts.