Avangrid to harness AI for the grid

Iberdrola subsidiary Avangrid has assembled a seven-member team to develop AI and machine learning-based systems for grid management.

The seven-member ‘Data science and analytics’ team drawn from diverse fields including astrophysics, healthcare and finance is charged with creating three different systems, Predictive Health Analytics, GeoMesh and HealthAI.

Each will take existing data from Avangrid companies’ electricity grids and analyse it to forecast future performance of the grid, determine the condition of grid equipment or target at-risk locations for inspections and investment.

Ultimately, this should lead to increased reliability for the 2.31 million customers served by these companies, i.e. Central Maine Power, New York State Electric & Gas, Rochester Gas and Electric and United Illuminating.

Have you read?
AI to fundamentally change current systems in the energy sector – experts
Closing the digital skills gap is vital for Green Deal success says Van Stiphout

“We’re reimagining what’s possible for a utility when it comes to data science and analytics,” said Pedro Azagra, Avangrid CEO.

“Traditionally, we’ve partnered with third parties to integrate this type of cutting-edge technology into our business. Now, we have the talent in-house to create machine learning models that Avangrid will own.”

The ‘Predictive health analytics’ system is focused on taking a proactive approach to determine the condition of substation equipment such as circuit breakers and enable replacement before an outage occurs.

Traditionally, equipment is replaced primarily based on age or if it malfunctions and causes an outage. ‘Predictive health analytics’ will determine equipment’s overall health and life expectancy based on numerous factors, including age, frequency of use, and manufacturer and maintenance notifications.

The ‘GeoMesh’ project is mapping Avangrid’s service areas to identify the strengths and weaknesses of its networks to help forecast its performance during both blue-sky and storm scenarios.

To accomplish this, ‘GeoMesh’ breaks Avangrid’s service areas into small sections, for each of which predictions can be made based on data points, such as average wind speed, precipitation type and amount, outage history and reason, population and density of tree limbs and other vegetation.

The ‘HealthAI’ project is analysing Avangrid’s existing millions of high-resolution photos of its street-level distribution system to identify the assets visible in them and to catalogue their health, with a view to increasing awareness of their health and to help identify areas of concern.

Currently, the AI system is being trained to correctly identify the grid equipment in the photos and then it is intended to learn to analyse and determine the health of that equipment, e.g. if a cross arm is broken or if a wire is sagging.

With ownership of the AI systems, Avangrid anticipates their continual improvement, improving their cost-effectiveness.