How AI Is transforming utility resilience

As extreme weather events grow in frequency and intensity, utilities face mounting pressure to strengthen infrastructure and demonstrate that resilience investments are working. Yet measuring resilience—especially in a way that accounts for the complexity of weather impacts—remains a challenge.

Utilities have long tracked outages and system performance but measuring how well a system withstands and recovers from severe weather requires a different approach. AI offers a new way forward by identifying not just individual weather variables, but the combinations of conditions that most often lead to service disruptions.

A smarter way to understand weather impacts

Recent applications of unsupervised machine learning—specifically self-organizing maps—have enabled a more detailed understanding of how weather affects grid performance. By clustering hourly weather data and correlating it with outage records, AI can reveal patterns that previously were anecdotal or intuitive.

For example, while high winds alone may not explain widespread outages, AI can quantify the extent to which wind combined with saturated soil from prior rainfall significantly increases the likelihood of tree-related damage. This insight helps utilities better understand the conditions that truly drive outages.

Defining resilience through exposure and performance

From recent analysis, two resilience metrics have emerged as valuable tools for utilities:

• Preparedness metric: Characterizes the degree to which resilience attributes are present in a system (e.g., floodwalls at substations) and helps identify investment priorities based on local climate hazards. For instance, some regions will have a greater need for flood preparedness than others, and the preparedness metric can take the local climatic needs of a region into account in the context of which resilience attributes are present.
• Performance metric:Provides a quantitative view of how a system performs during severe weather events, relative to the severity and frequency of those events. It assesses outage frequency and duration for each weather type—factoring in event characteristics, number of customers affected, and outage length.

These metrics offer a structured way to benchmark systems, track improvements, and guide resilience investments with greater precision.

Looking ahead: Predictive power and planning

While current models rely on historical weather data, the same techniques can be applied to future modeling and projections. This enables utilities to estimate how outage patterns may change over time. These insights can support better proactive investment and long-term planning.

By using AI to analyze the relationship between weather and outages, utilities gain a valuable tool for shaping resilience strategies. The framework is flexible, scalable, and adaptable, offering a strong foundation for broader industry use.