Sky over a solar panel farm

AI SOLAR POWER PREDICTION CHALLENGE

Powered by Shell, Microsoft, NVIDIA, OpenACC, IIT Bombay and IIT Madras

Meet the winners of 2021 Shell.ai Hackathon: AI Solar Power Prediction Challenge

Shell.ai hackathon’s ambition is to create AI-based solutions to make energy more sustainable and affordable. In this edition, we asked the participants to predict the percentage of total cloud coverage and Global Horizontal Irradiance (GHI) over a solar farm to better forecast its power production.

The winners rose to the challenge and developed solutions that were accurate, innovative and scalable.

Meet the bright minds behind the winning solutions and read about their experience of the hackathon. Congratulations to the winners! 

General Edition Winners

1st Prize

Dr. Sukanta Basu

Dr. Sukanta Basu

Associate Professor

Delft University of Technology

I’ve been doing renewable energy related research for last almost 20 years. I’ve also had experience with machine learning going back to 1999, but for the last two years I’ve spent a lot of time understanding the new deep learning tools that could really help my research. This hackathon was a challenging one, full of excitement (and some frustration) – I had my ups and downs working on the solution. But if you know the domain reasonably well, I think machine learning can do miracles for you. I guess all these years spent studying atmospheric physics actually paid off.

2nd prize

Ricardo Lara (Left) and Roderick Perez (RIght)

esk.AI

Ricardo Lara, PhD candidate in petroleum engineering at The University of Texas at Austin (left)

Roderick Perez, physical engineer, Data Science student at University of Vienna (right)

Ricardo: This hackathon has been a great networking experience. Me and Roderick never met in person. We connected in social media and decided to form a team. It was stimulating to discuss our ideas and get feedback from Shell and NVIDIA mentors, get inspired by these conversations and put new ideas into action.

Roderick: Both of us have background in oil-and-gas engineering and we are both passionate about access to energy. For me, it was an opportunity to step into the world of renewable energy and prove to myself I can be a part of a solution to the world’s current climate challenges.

3rd prize

Akshat Gupta (Left) and Sumit Yadav (Right)

peaceHai

Akshat Gupta, data scientist (left)

Sumit Yadav, data scientist (right)

Sumit: We’ve been working together in similar competitions for the past two years and got very intrigued by this challenge, as we realise forecasting energy production in solar farms and its impact on supply and demand is a valid problem in the energy space.
We are motivated by a vision of making a real change, so we are thinking about publishing a paper and releasing our source code so the whole community can benefit from our solution.

Akshat: We loved the competitive nature of this hackathon. Monitoring the leaderboard and continuously improving the accuracy of our algorithm to get ahead of other teams kept us going. But it was about more than good scores. We paid extra attention to our model being scalable and deployable, so real people can use the technology to make their lives better.

Special Start-up Edition Winners

Meson Labs

Ashish Sharma (Left) and Ritu Sinha (Right)

Ashish Sharma, data scientist at Meson Labs (left)

Ritu Sinha, Founder- Director and Head of Research at Meson Labs (right)

Ashish: At Meson Labs we work mostly on AI machine learning solutions for three sectors: education, healthcare, and climate technology. We look at applications of technology that make a sustainable impact, so the hackathon sparked our interest immediately.

Personally, I got interested in renewable energies, as my father who had worked lifelong designing and commissioning steel plants, coal powered blast furnaces, moved into solar power and renewables. It’s exciting to be able to use the skillset and talent we have at Meson-Labs to create something that can have a long-term positive impact on the climate. We built our models with deployment and scalability in mind.

Akira.Insights

Akhil George, data scientist (right) and full team (left)

Akhil George, data scientist (right) and full team (left)

Akhil: Akira.Insights is a small start-up that provides data science and artificial intelligence solutions. When we saw the hackathon’s problem statement, we decided to challenge ourselves and see if we can crack it and how far we will go.

We invested a lot of time and effort in our solution and at the end of the day it was extremely rewarding, as we believe that our products can directly help companies and society. I have experienced the intermittency of solar power first-hand, as my home is fully solar-powered. It’s a great feeling to work on something that touches you in such a direct way.

Access to NVIDIA’s GPU cluster was a game changer for us. The computing power made it fast and easy to try out different solutions. Personally and professionally, I feel this hackathon pushed me to the next level.

About the Shell.ai Hackathon for Sustainable and Affordable Energy

Tackling climate change is an urgent challenge. Shell is transforming to become a net-zero emissions energy business by 2050, in step with society and our customers. We are exploring new opportunities to provide more low-carbon energy such as biofuels, hydrogen, charging for electric vehicles and electricity generated by solar and wind power. To achieve this, we are working collaboratively on unproven, early-stage ideas with the potential to impact the future of energy.

Solar power is one of the fastest growing renewable energy sources. However, there is a major challenge associated with solar power production, and that is its intermittency caused by variable weather conditions. Of course, we know the sun is not going to shine at night. Predicting solar power output during the day is key for grid operators to balance energy generation and consumption.

We believe that bright minds passionate about AI can creatively and effectively address this problem and contribute to increasing the share of renewable energy in the electricity mix. That is why Shell, NVIDIA and Microsoft, with support from SINE (IIT Bombay) and NIRMAAN (IIT Madras), launched AI Solar Power Prediction Challenge that focused on AI for predicting solar power output using historical sky camera images and weather data. The winners of the hackathon got awarded with either monetary prizes (6,000 USD combined) or funding to further develop their Proof of Concept.

CHALLENGE FORMAT

This edition of Shell.ai hackathon will run in two tracks, providing the opportunity for both individual teams and start-ups an opportunity to compete.

 

General Edition (Track 1) Special Start-up Edition (Track 2)
Eligibility Open to all individuals and teams of up to 4 members (except Shell, Microsoft and NVIDIA employees and contractors) Open to registered start-ups only (minimum 2 members)
Bootcamp and mentoring Shortlisted 20 teams Shortlisted 10 teams
Winners and awards
  • 3 winners
    • 1st prize: 2,500 USD
    • 2nd prize: 2,000 USD
    • 3rd prize: 1,500 USD
  • Maximum 3 winners
  • A chance to receive Proof of Concept funding and develop their products with Shell GameChanger (up to 150k USD), Microsoft Incubation Hub and potentially SINE (IIT-B) and NIRMAAN (IIT-M) (subject to Shell’s and our partners’ criteria and regulations)

The challenge will consist of three levels:

Level 1

Level 1 – Predict cloud coverage (October 7 – November 7, open for general and start-up edition)

Cloud coverage can cast over the solar farms in a few minutes and significantly bring down the power production. In this level, participants are invited to innovatively predict the percentage of total cloud coverage for the next upcoming intervals using the available weather and sky camera data. The hackathon teams can decide their own methodology to predict the cloud coverage in the sky. The submitted algorithm solutions will be evaluated based on the accuracy of the predictions. The 30 teams (20 from the general edition and 10 from the special start-up edition) with the highest scores will be shortlisted for the next level.

Level 2

Level 2 – Predict Global Horizontal Irradiance (November 15 – November 30, open for general and start-up edition)

In Level 2 of the challenge the participants will be asked to predict GHI (Global Horizontal Irradiance) for any 2-hour interval of the day, subject to the availability of a minimum 2-hour window of prior data, leading to the start time of the prediction window

In this phase, we are increasing the complexity by two ways:

  1. You are expected to predict GHI (Global Horizontal Irradiance) for any 2-hour interval of the day, subject to the availability of a minimum 2-hour window of prior data, leading to the start time of the prediction window.
  2. The prediction interval is now at 10 mins, resulting in 12 predictions over the 2-hour window.

In this round, we will also provide additional information relating to sky camera images. In addition to the raw sky images, we will provide scaled images that truly represent the pixel size.

For more information, please view Shell.ai Hackathon Level 2 Problem statement.

  • The winners will be evaluated based on private leaderboard ranking, as well as qualitative factors that will be shared with participants after the start of Level 2.
  • The metric to evaluate your submission will be Mean Absolute Error (MAE).
  • Hackathon committee shall have final authority in selecting applications as per the suitability and nature of the proposals.
  • Hackathon committee shall have final authority in selecting winning teams as per the criteria mentioned in the FAQ section.

Level 3

Level 3 – Solutions to reality (for the start-up edition only)

Maximum 3 selected winning start-ups with the best pitch and solution from level 2 will be awarded funding for developing their proof of concept and an opportunity to incubate their solutions with Shell GameChanger, Microsoft Incubation Hub and potentially with SINE (IIT-B) and NIRMAAN (IIT-M) . We will help set you up with the capital and technology you need to bring your solution to life.

Frequently Asked Questions

Logos of partners

Our partners

FIND OUT MORE

Digital and AI Competitions

Are you passionate about novel solutions that could impact the energy future? Sign up for our digital and AI competitions.

Shell GameChanger

We work with start-ups on unproven early-stage ideas with the potential to impact the future of energy.

Solar

Solar energy is one of the fastest growing renewable energy sources. As photovoltaic technology advances and production costs decrease, Shell is incorporating solar energy into its portfolio as customer choice drives the shift toward decentralised electricity.

1st edition of Shell.ai Hackathon for Sustainable and Affordable Energy

Around 5000 contestants signed up in the 1st edition of Shell.ai hackathon to compete in the Windfarm Layout Optimisation coding challenge. 14 teams have made the finalist shortlist. 8 have emerged as the winning teams. Watch the awards ceremony video below.