How can we put restoration history to work for a 30-second decision?

By Jim Nowak, Contributor

By inputting the proper parameters, artificial intelligence could help a utility respond to major events and daily challenges based on, for example, the history of average restoration time per broken pole. How can utilities take advantage of that information and put it into what the All Hazards Consortium calls “30-second decision making”?

Merriam-Webster defines artificial intelligence, or AI, as “a branch of computer science dealing with the simulation of intelligent behavior in computers.” AI technology is supposed to help computers learn and complete tasks that humans would ordinarily undertake. With AI technologies in place, there are natural language processing capabilities that carry out calculations as opposed to conventional software employing decision trees that then require humans to perform work manually. Examples of AI technology include chat bots, speech recognition, biometrics, image and video analysis and machine learning. AI systems learn from analyzing large data sets.

For utilities, when it comes to artificial intelligence, there’s value in integrating AI with outage management systems (OMS) and resource management platforms, which track line crews, equipment and damage. With these storehouses of data to draw from, an integrated AI tool could model a scenario in which a utility faces, for example, a wind storm impacting 50,000 customers. AI could then automatically look at historical data for a similar event, estimate a number of resources for a safe and efficient restoration, and provide an ETR. If the ETR is not acceptable, the AI technology could adjust the number of resources based on a set ETR, say 48 hours for less than 100,000 impacted customers. AI then could look at existing resource availability and determine the delta allowing leadership (i.e., a 30-second decision) to approve additional resources and reach out to neighboring utilities and contractors to meet the need. AI could go a step further and, based on history, provide an estimate for the number of:  broken poles, damaged transformers and downed wires; required crews and their makeup; damage assessors and tree crews.

A lot of the building blocks to make the above scenario a reality already exist with technology systems at U.S. utilities. For instance, a typical OMS can provide a utility manager data via smart meters, SCADA and customer calls. If integrated with other systems, an AI system could offer options to improve and update that ETR with a link to an automated crew-management solution that provides crew makeup (i.e., number of employees, types of vehicles, etc.) and location. Add real-time damage assessment and input from local officials about streets blocked by storm damage, and AI could deliver routes for crews to take as well as a priority list of areas for debris removal. An AI-powered system like this would save utility managers from manually adjusting data and instead deliver options they could quickly choose.

AI’s role in constructing an outage history

Not long ago when unfamiliar events struck a service territory, utility managers would ask someone in the company who had experienced a similar outage – sometimes years before – to recall how many resources were deployed and in what manner. More recently, utilities began developing prediction models that capture historical data related to weather, damage and customers. That new approach has improved restoration forecasting, but there are some potential risks with using historical data. Here’s one: Today’s customer has a heightened sense of awareness to outages and ETRs, so utilities have been hardening circuits, employing bigger poles and aggressively managing vegetation. With that in mind, utility managers should consider outage history prior to their utility’s infrastructure hardening as a suggested solution, which is not unlike the way managers must see AI. Leaders have to make the final call, while considering many factors. Over time, as hardening becomes the norm, outage history will better account for improvements and thereby sharpen forecasting.

With AI, the utility industry can harness myriad factors to construct an outage history, which managers can use to see a storm model and build a response. An outage history developed with AI tools might analyze the time of year of the last major event, including the level of foliage on trees and how many days prior to that event the last rain occurred. It would calculate the anticipated amount of rainfall, since rain levels affect root systems making trees easier to topple in high winds. The AI-built model might consider wind direction, too. Since trees tend to grow in a direction that helps them withstand prevailing winds, an AI tool could input a wind speed from a new direction and perhaps determine if the utility would be dealing with increased limb damage and fallen trees.

Today, there are multiple technologies to capture data, predict outages and respond to an event. The key is corralling them to work as one. IBM has a program called Deep Thunder, which has as its goal improving weather forecasting. The Deep Thunder project mined historical data on the damage caused to, for example, power lines. Deep Thunder took that data and combined it with a forecast for a very focused locality, which IBM hoped would help utilities estimate the number of restoration crews needed and where to dispatch them. For Deep Thunder to work effectively, researchers say it’s important to have “historical impact and infrastructure data from the clients.” Imagine a tool like Deep Thunder and the data from SCADA systems, customer calls and smart meters combined with mobile damage assessment tools. With a hyper-local, accurate forecast, plus real-time outage and damage reports, utilities could begin alerting the correct number of native and contract crews before a storm and dispatch and release them with far greater clarity than ever before.

Ultimately, AI isn’t about supplanting critical, human decision-making. Instead, AI tools bring together the history, resources and forecasts that humans would in years past try and complete manually. AI assembles scenarios far faster than humans can do for themselves. And utility managers quickly gain a perspective and confidence to make a 30-second decision.

Jim Nowak retired as manager of emergency restoration planning for AEP in 2014. He capped his 37-year career with AEP by directing the utility’s distribution emergency restoration plans for all seven of the company’s operating units, spanning 11 states. He was one of the original co-chairs for Edison Electric Institute’s (EEI) Mutual Assistance Committee and National Mutual Assistance Resource Team and a member of EEI’s National Response Event (NRE) governance and exercise sub-committees. He is senior director of Operations, Product and Services for ARCOS LLC. Contact him at

Previous articleRocky Mountain Power, partners launch energy storage grid project
Next articleAiming for the X: How a new experience-focused business paradigm is shifting the balance of power in electricity markets
The Clarion Energy Content Team is made up of editors from various publications, including POWERGRID International, Power Engineering, Renewable Energy World, Hydro Review, Smart Energy International, and Power Engineering International. Contact the content lead for this publication at

No posts to display