Space Apps Challenge 2017

Solar panels are dropping in price each year. It is only a question of time that people will start installing them on every free spot around us. Solar panels should be connected to the global power network to provide support for traditional energy sources. And eventually replace them entirely, combined with other “green” renewable sources of energy. Solar panels are more efficient when located closer to the consumer to reduce energy losses on the long wires. Solar panels should not occupy useful areas like agricultural fields or other human activity zones.

SolarCity project uses neural networks for satellite images processing to find roofs suitable for solar panel installation within a selected neighborhood, town or a city. By estimating the effective roofs surface area, our software calculates solar system parameters: energy production by month, number of roofs, roof total surface area, solar panel total surface area, average energy production per roof. The project also visualizes roofs with solar panels on the satellite map.

SolarCity project does not aim on calculating economical relevance, but rather provides information needed to create an investment or subsidizing strategies for the municipal government or planning future energy system growth.

https://github.com/azati/nasa2017

License: Apache License

Library references:
http://deeplearning.net/software/theano/
https://www.openstreetmap.org/#map=5/51.500/-0.100