genitstat_TheTalkingTrees

We extract rooftop information from geographical data and satellite images to determines the potential yields of solar panels

Comment

Inspiration

Due to the supply chain limitation; skill shortages and many other factors, there are many unnecessary troubles when interested customers want to install solar panels. This project aims to help Germany utilize its solar panel resources as efficiently as possible. Our project presents a method to address this problem with the help of AI. We propose a pipeline to create a dataset to determine the rooftop properties of buildings using satellite images. By combining this information with sun radiation data and tilt angles, we estimate solar panels' electrical potential and efficiency on these roofs, providing a good estimation for corporation or the government to decide on which area should solar panels be installed for the best energy yields.

What it does

It estimates solar panels' electrical potential and efficiency on the rooftops of big cities in Germany.

How we built it

  1. Data Collection and Preprocessing: We collect satellite images and OpenStreetMap data to create images of roof shapes in Germany. Additionally, we gather sun radiation data to be used in the estimation process.

  2. Roof Shape Detection: We train a CNN model to classify roof shapes in the collected images. The detected roof shapes serve as input for the azimuth angle detection step.

  3. Azimuth Angle Detection: We use CV techniques to determine the azimuth angles of the detected roof shapes. This information is essential for calculating the solar energy potential.

  4. Energy Potential Estimation: With the roof shape, azimuth angle, sun radiation data, and tilt angle, we estimate the electrical potential and efficiency of solar panels on these roofs.

Challenges we ran into

  1. Too many buildings to process.
  2. Coordinates mismatch between OSM and Google Maps
  3. Many unusual rooftops

Accomplishments that we're proud of

  1. We did it within less than 2 days
  2. We provide a way to help proliferate sustainable energy in Germany. A lazy method was implemented for everyone to estimate the potential energy of their own house
  3. We can sleep less than 2 hours a day (for 1 day only)

What we learned

  1. Geographical data and some insights into the solar business
  2. AI for satellite data

What's next for genitstat_TheTalkingTrees

  1. The pitch at 1pm, the 2nd pitch at 3pm, the party at 7pm.
  2. Improve the AI algorithm to create a rooftop dataset for the whole Germany.

Built With

Share this project:

Updates