Random Forest Approach to Resilience-Oriented Infrastructure Post-Disruption Repair Optimization

Resilient critical infrastructure systems should be able to quickly recover from disruptions to ensure sustained service deliveries. To speed up the recovery process, optimal recovery strategy is crucial for better situational awareness. Current research on sequencing the repair of disrupted power system components either use a generic network model as an approximation or rely on rule-based methods which might not produce recovery strategies that minimize resilience loss. We propose a random forest-based recovery sequencing optimization strategy based on stochastic disruption simulation data obtained from realistic system simulations. The strategy is shown to be more effective in limiting resilience loss compared to existing methods when tested with six-line and ten-line disruption scenarios from the IEEE 39 test system. Having a fast and effective repair sequencing strategy with the explicit goal of limiting resilience loss could provide valuable service to operators responding to disruptions in power systems.