Solar Capacity Estimation from Aerial Imagery

The amount of solar panels is increasing rapidly. Current solar power capacity estimation relies on solar panel owners reporting capacity, which is a slow process. This report presents a method to predict the power capacity of solar panels using the area from two dimensional satellite images. The goal is to show the correlation between these two variables and to quantify it. A data set from North Carolina Sustainable Energy Association, NCSEA, was used were solar panels were reported with a capacity and an approximate location for solar cells in North Carolina, USA. A polygon was created for the solar panels' boundaries using satellite images and from these polygons an area was calculated using a Matlab script. After creating a dataset with both capacity and area, a linear correlation was calculated. A linear regression analysis was run in order to calculate confidence and prediction intervals. The results show a strong correlation between solar panel area in satellite images and solar panel capacity. A general linear correlation was quantified, alongside with a refined correlation for solar panels with an area less than or equal to 200 m.