Power System Resilience under Natural Disasters

Abstract

Power systems are not likely to remain unscathed by natural disasters such as earthquake, hurricanes, ice storms, as evident from the recent Hurricane Harvey and Hurricane Irma. The outages will last days or even weeks because of the amount of damaged components. And the impacts are aecting the economies, public health and communities especially those that are already facing challenges. This motivates us to study methods of improving resilience in both operational stage and planning stage. We believe this is an interdisciplinary research from several aspects, 1) There has been no consensus on the definition of power system resilience under natural disasters. And in fact, this research direction only becomes hot in recent 4 or 5 years. However, the concept of infrastructure resilience has been prevailing and well-studied in civil engineering. After summarizing previous efforts on defining and quantifying of resilience including those adapted to power systems, we base our work on the resilient measure derived from operability trajectory and develop an equivalent measure of harm that has clearer power system meanings. 2) The knowledge of power systems guides us to focus on electricity distribution systems, where we believe the resilience has more potential for improvement. We start with the case of fully automated radial distribution network, and then move on to partially automated radial distribution network and finally find a way to handle the uncertainties in repair time. After consulting with industry experts, we relax certain operational constraints to make the problems (slightly but enough) easier to solve without compromising their practicality in field. Built upon the operation problems, we formulate the quantification and assessment of resilience in the planning stage, which will help electric utilities decide how best to spread the budget to improve the resilience. 3) Unfortunately, none of the problems described above are easy to solve in terms of the computational complexity. In particular, the operational problems might need to be solved in real time repeatedly and MILP formulations, though straightforward, are too slow in practice. We adopt the settings of scheduling theory and propose the first of its kind, soft precedence constraints, to model the relaxed load flow equations in radial distribution networks. And for the assessment of resilience in the planning stage, we simplify the operational problem by using a single crew approximation with only a constant away from optimal. This allows us to reformulate the distribution systems hardening problem into a combinatorial optimization with the flavor of the multiple knapsack problem. To summarize, this research aims to develop good algorithms and heuristics for problems under the framework of power system resilience adapted from the concept of infrastructure resilience.