In 2003, Mother Nature turned off the lights on the East Coast. The reason: a short circuit a hot summer day caused on by a chance encounter between an overgrown tree branch and a sagging power line. The problem quickly cascaded through the system, triggering the biggest blackout in North American history.
The outage left 50 million people in the U.S. and Canada without power and by some estimates cost more than $6 billion. But the truth is, most people don’t give much thought to our electrical grid until something goes wrong. Most, but not all. Behind the scenes, a dedicated group of technicians is working hard to find efficient ways to monitor the health of the power network and prevent similar blackouts from happening.
The job isn’t easy. It can involve hiking through rugged terrain with cameras and binoculars and looking for leafy tentacles that may be snaking dangerously close to high-voltage lines.
But the ground crews are now able to call for air support. Field service technicians from GE’s Grid Solutions business can bring in flying drones equipped with digital sensors and cameras that can take detailed images from the air and relay them instantly to a cloud-based platform for analysis. The company behind this platform is Avitas Systems, a Boston-based GE venture launched in 2017 that applies robotics, artificial intelligence and risk-based predictive analytics to industrial inspections.
Replacing the manual visual surveys with data such as images and video allows for a more precise inspection, and it also enables the customer to build a historical archive of evidence for use in predictive decision-making. “The software has artificial intelligence to predict where there could be a problem,” says Chantal Robillard, manager of asset life-cycle management services for GE Grid Solutions.
Of particular interest to grid operators: Avitas Systems is creating AI-powered modules for vegetation encroachment to detect the type of tree near a given power line — then use information on that species to predict its growth rate based on weather patterns and seasonality. The equipment owner can then make an informed decision about whether and when to trim the tree — before it causes any problems.
In addition to vegetation management, AI can differentiate between foreign objects, like birds’ nests, and grid components such as insulators and spacers. The platform can also identify abnormalities like corrosion, cracks and thermal spikes.
Over time, the AI learns what is typical for a specific power line and will pick up on anomalies it detects in the image data. For example, if a drone is monitoring an area in Southern California, the AI will predict the growth patterns of the local palm trees, and the system will send out alerts long before there is a danger of a frond touching the line.
But there’s more to the story. The images are uploaded along with their accompanying analysis into GE’s asset performance management (APM) software, complementing a myriad of additional information, such as oil analysis and measurement from online condition-monitoring systems, for a complete view of the power grid equipment’s health. With the pooled data, GE APM can then build a digital twin of the power system — its virtual model — and use advanced analytics to predict what might happen in the future. For example, the program can estimate the ideal time for a repair based on how likely the equipment is to fail and when, how old the equipment is, and how important it is to the grid system as a whole. This allows GE’s Grid Solutions’ services experts and their customers to develop the best maintenance strategy based on the customer’s priorities for managing risk, improving reliability and reducing costs.
The system doesn’t rely only on drones. Stationary high-definition and infrared cameras can perform visual inspections that technicians traditionally perform. For example, an infrared-equipped camera installed inside an electrical substation — the physical heart of the grid, where the voltage levels are ramped up or down — can look for hot spots that could indicate a transformer is prone to overheating. Because the camera is always on, it can pick up sudden and potentially damaging temperature spikes that might otherwise go undetected until the next maintenance visit. Says Robillard, “We believe this is the future.”