Measuring Solar’s True Power

Energy_team-2-sized
tags: none | See all tags

Programs like Data+ give Duke students the chance to get “under the hood” and use data science to solve real problems.

Solar power is gaining traction in the United States, but by how much? One group of iiD students has created a sophisticated dataset that can train machines to identify rooftop solar panels in satellite images. In preliminary results, a machine-learning algorithm they developed was able to identify rooftop panels with 90 percent accuracy. These data-powered tools could one day provide accurate estimates of the country’s—and world’s—solar capacity. This would help energy companies, urban planners, and policymakers make better-informed decisions that will take us closer to affordable, accessible clean energy.