Can we clean energy data with Neural Nets?

The human brain in amazing. It makes sense of surroundings in ways we don’t always appreciate; for example, our ability to navigate from point A to point B. Along the way we combine systems across our minds. At any millisecond, billions of nerve cells inside our skulls spark and react.

The magic about all of this is that it serves as a blueprint to solve problems all around us. For instance, innovators like Professors Tom Caudell and Andrea Mammoli of University of New Mexico correlated how our visual systems can perceive distance and speed based on cues around us such as differences in how fast objects like poles and mountains move away from us while we drive. This information can be stored in short-term memory and compared through short intervals of time to update our sense of motion and placement in our world. Together with EPRI, they have written about the use computational models that mimic this mental process to make data more accurate without the inconvenience of having workers drive to each pole to capture its exact location coordinates.

Humanity needs to gather data for exact locations and status of power lines and other electrical assets that keep our favorite fast food restaurants producing the meals we crave. It takes a lot of manpower to drive out to collect the information and update it to maintain accuracy. Professors Tom Caudell and Andrea Mammoli suggest we can using images from Google Street View to detect where the poles are rather than drive to every one of them.

Their research indicates that Neural Networks, brain-inspired of computational models of our visual system, can detect types of location of energy transmission assets in an image and then link this to the geographical place where the picture was taken, comparing it with previous and subsequent imagery of the same scene to pinpoint assets location within 1 meter. To do this, the software mimics interplay between Ventral (‘What’) and Dorsal (‘Where’) systems in our brains.

This is just one of the innovate ways that deep learning, neural networks and artificial intelligence can be used to improve accurate data for energy grids.

Article by Micah Tinklepaugh and John Simmins