The increase in power outages associated with adverse weather events in the United States has had significant economic ramifications. Climate change has exacerbated the occurrence of extreme weather events such as hurricanes, heat waves, wildfires, and snowstorms. This, combined with increased power demands, is now regularly causing billions of dollars’ worth of damage due to resultant power grid disruption. Between 2003 and 2012, 679 power outages were caused, which affected at least 50,000 customers, and up to 90% of these failures are attributed to disruptions in the distribution systems [1]. Some of the more prominent occurrences in recent memory include Hurricane Sandy in 2012, which left 8.5 million people without power [1]; the 2021 Texas winter storm, which affected a million residents [2,3]; and the 2017 Hurricane Irma in Puerto Rico, which affected power distribution to 6.7 million customers [4]. The economic costs associated with these disasters run into the hundreds of billions and, at the same time, lead to significant societal damage. For instance, the Texas winter storm resulted in multiple deaths because electric heaters and medical equipment could not run without power, leaving vulnerable people exposed to extreme cold and/or without critical life support [2,3]. One reason why power systems are particularly vulnerable to extreme weather events is that they are mainly designed and optimized for normal weather conditions and are not equipped to handle less common extreme weather events. While this is economically viable in the short term, it exposes the power grid to catastrophic failures that are costlier in the long run. Another essential dynamic to consider is that disasters that lead to shocks and stresses are often connected, making the risk two-fold. As such, the risk is multidimensional and self-perpetuating. For example, higher temperatures due to climate change result in higher power demands, therefore increasing the load on the power system. At the same time, there is a higher risk of adverse weather events, which means that when disaster strikes, it finds an already compromised power grid that is easier to knock down. It is, therefore, paramount that utility companies strengthen their systems’ readiness for extreme events. During the past decade, there has been much focus in the literature on the resilience of the electric power grid against large-scale disturbances, mainly natural disasters. Figure 1 illustrates a keyword co-occurrence map of the recent publications on power grid resilience.

Although more attention has been given recently to making the grid more resilient, a standard definition for power system resilience does not exist in the literature, and while different researchers are generally converging to the same sets of criteria, various definitions are being presented and adopted. A common theme for a resilient power system is one that has the capacity to withstand disturbances and continue to deliver energy to customers [5]. Some have further specified that resilience is defined against extraordinary events, referred to as high impact low frequency (HILP) events [6,7]. Many studies in literature specifically focus on the resilience of the power distribution grid [8,9,10,11,12,13,14,15,16,17] as it is more closely related to customers. For this, most use the definition from the US Federal Energy Regulatory Commission (FERC) [18], in which resilience is defined as “the ability to withstand and reduce the magnitude and/or duration of disruptive events, which includes the capability to anticipate, absorb, adapt to, and/or rapidly recover from such an event”. Here, being able to withstand an event is further broken down into the ability to absorb its impacts and adapt to the changes caused by it. Moreover, the ability to anticipate the event is considered as yet another dimension of resilience. The definition by CIGRE working group C4.47 introduces additional dimensions listed as anticipation, preparation, absorption, sustainment of critically operated systems, swift recovery, adaptation, and implementation of the learned lessons from previous events [19]. In a more general context, the National Infrastructure Advisory Council (NIAC) [13] has listed four attributes of a resilient system, which are robustness, resourcefulness, rapidity, and adaptability: Robustness is the capacity of the system to withstand the impacts of an event without suffering severe performance deterioration or loss. Resourcefulness is the ability to initiate a solution by mobilizing resources such as information, capital, technology, and manpower. Rapidity is the ability to restore functionality and mitigate losses quickly. Finally, adaptability is the capability to learn new information from a HILP event and dispatch new solutions to increase the dimensions of robustness, resourcefulness, and recovery before the next event.

Strategies to improve resilience can be divided based on the timeline of the HILP event as preventive measures, which are adopted before the onset of the event; corrective measures, implemented during the course of the event; and restorative measures, which are employed after the event has run its course. Although these are decoupled from one another based on the timeline of their deployment, they can still impact the success rate of their counterparts. For instance, grid-hardening techniques that are normally implemented long before the onset of an event can lead to higher chances of success for a risk-based grid dispatch as the event is unfolding. Depending on the type and characteristics of the HILP event, some of these mitigation measures may be prioritized over others. For instance, solutions against short-duration events such as earthquakes are mainly focused on prevention and restoration, whereas for events that run a longer course, e.g., most weather-induced natural disasters, all mitigation aspects are considered. Figure 2 illustrates some of the general resilience solutions that can be applied in power systems.

Before the onset of an upcoming HILP event, resilience can be achieved by hardening the infrastructure, e.g., using stronger overhead line poles/towers or reinforcing them by installing guy wires, elevating substations in flood zones, waterproofing control rooms, or replacing overhead conductors with underground cables. Pre-emptive strategies can also be adopted by reducing the chances of disturbances and/or outages, for instance by increasing the distances between phase conductors to reduce the chances of flashover due to conductor slapping or vegetation management to prevent line contract with tree branches. Lastly, redundant designs can be employed to increase the resilience of the power grid. Some examples could be installing multiple circuits in parallel, using main and backup transformer units, or expanding the generation reserve capacity of the system by deploying additional centralized or distributed energy resources. These solutions are naturally costly and take place over a longer horizon.

During the course of the event, however, the focus must be shifted to adaptive operation. The availability of lines, transformers, and/or generation resources may be impacted by the event, in which case, energy dispatch needs to adjust by supplying power via more reliable resources and/or routing power through more secure lines and substations. Load curtailment, either voluntary in the form of demand response or involuntary in the form of load shedding, may also be adopted to ensure the stability of the power system. Due to the possible uncertainties associated with the nature of most HILP events, i.e., their intensity, scope, or trajectory, the corrective energy dispatch must be risk-based in order to protect the grid against all likely scenarios. Finally, during the restoration phase, it is assumed that the event has ended and, as such, the focus is on using alternate energy sources and alternate routes to provide power to the outage area, i.e., healthy parts of the grid that are left de-energized due to the event. Various network reconfiguration strategies may be used here to allow for higher connectivity.

2. Resilience Metrics