Optimized recovery of damaged electrical power grids

Abstract

This thesis formulates and solves a mixed-integer program to plan the recovery of an electrical power transmission grid that has been damaged by a natural disaster or terrorist attack. The damage can be extensive and recovery can take weeks or months. An efficient recovery plan that maximizes the utilization of repair resources can help ensure swift restoration of services. The network recovery-planning model is implemented in GAMS (General Algebraic Modeling System) and uses CPLEX as the solver. An electrical grid based on IEEE's 300-bus transmission network is used for testing. To simulate varying degrees of damage to the network, we choose up to 20% of the grid's lines, buses and transformers to be placed out of service. Based on the availability of repair resources, the repair-time horizon and penalties for unserved demand, the model produces a repair schedule that minimizes the cost of power shed. We demonstrate that for a network with up to 8% of its components damaged, the model can produce an optimal recovery plan within 20 minutes on a 2 GHz personal computer. For our largest test-case with 20% of network components damaged, the recovery plan is within 7% of optimal after 1 hour of solver time.