Avangrid launches AI team to leverage data, anticipate future grid challenges

Avangrid is launching an in-house artificial intelligence team to leverage existing data to improve the reliability of the grid and position the company to successfully tackle future changes in the electric system, a top operational official at the company told S&P Global Commodity Insights.

"We really see these use cases and these analytics as the beginning of what I think could be a really transformational use of data at Avangrid," Mark Waclawiak, senior manager of operational performance at Avangrid, said.

The AI team will take a large amount of data that Avangrid already collects about assets, historical performance and outages and combine it with weather and geospatial data to forecast the future performance of the system and determine where to invest in the grid, Waclawiak said.

Avangrid's Data Science and Analytics team is made up of seven data scientists, engineers and analysts, the company said in a statement. The team is creating three main AI systems: Predictive Health Analytics, GeoMesh and HealthAI.

The tools will be used to improve reliability at Avangrid subsidiaries Central Maine Power, New York State Electric & Gas, Rochester Gas and Electric and United Illuminating, the statement said.

AI systems

Predictive Health Analytics will use data like historical failure rates, maintenance notifications and inspection information to predict the lifespan of substation equipment like circuit breakers, and replace equipment before it fails, Waclawiak said.

"We see AI as a really effective way to bring all of that data together through the development of machine learning models or neural networks." Waclawiak said.

HealthAI will analyze Avangrid's existing millions of photos of poles, wires and grid equipment and categorize their health, the statement said.

GeoMesh will map Avangrid's service area to identify strengths and weaknesses of its electric networks and to forecast performance in different weather conditions, the statement said. GeoMesh will analyze data such as average wind speed, precipitation type and amount, outage history, population and density of vegetation, the statement said.

"What we wanted to do is combine different data sets to begin running localized data science models, AI models, to understand system performance in our territory of New York, Connecticut and Maine," Waclawiak said regarding GeoMesh.

GeoMesh will help Avangrid understand if there are areas that are more susceptible to higher winds or heatwaves or wet snow, Waclawiak said. Taking a localized look at how the system performs allows Avangrid to better tailor its investments or initiatives with grid hardening or automation," he said.

"That's kind of the thinking with these three projects, right, using a lot of the data that we have as a utility in much more advanced ways to really drive operational insights," Waclawiak said.

For instance, Avangrid has used the GeoMesh tool in the Ithaca, New York region to determine how tree trimming corresponds to performance as it relates to different weather conditions, Waclawiak said.

In a heavily wooded area, the analysis helps Avangrid decide whether it should pursue more tree trimming, use a type of wire that can handle tree contact, or use a device that provides an automated, momentary interruption to prevent a more sustained outage, he said.

Holistic view of risk

"So to me, it is a holistic view of risk of the electrical system that allows our operational folks, our planning folks to really make surgical investments in the system to maximize the reliability value of every dollar we spend," Waclawiak said.

Avangrid has traditionally partnered with third parties to integrate new technology into its business, the company said in the statement. But Avangrid decided to create its own AI to leverage its in-house expertise while also improving internal models and technology, Waclawiak said.

"In the long run this really does save money, it really increases efficiencies," Waclawiak said. As the tool ingests more data, the models improve, he said. "Because we are really the owners of it, the developers of it, we are also the beneficiaries of its consistent improvement," he said.

The program will also help Avangrid prepare for future changes in the grid, Waclawiak said. For instance, AI can help determine which regions are most likely to see increases in electrification, and Avangrid can then use this data to make the case to regulators that more grid capacity will be needed in these regions, he said.

Waclawiak argues that Avangrid's decision to develop an in-house AI team puts the company at the forefront of the industry. "We see these analytics as just the beginning of putting Avangrid in a position where we can say this is not just a poles and wires business, this is a technology company," he said.