Open Climate Fix is a new non-profit research and development lab, totally focused on reducing greenhouse gas emissions as rapidly as possible. Every part of the organisation is designed to maximise climate impact, such as our open and collaborative approach, our rapid prototyping, and our attention on finding scalable & practical solutions.
By using an open-source approach, we can draw upon a much larger pool of knowledge and skills than any individual company, so combining existing islands of knowledge and accelerating progress.
Our approach will be to search for ML (Machine Learning) problems where, if we solve a well-defined ML task, then there is likely to be a large climate impact. Then, for each of these challenges, we will:
- Collate & release data, and write software tools to make it super-easy for people to consume this data.
- Run a collaborative “global research project” where everyone from 16-year-olds to PhD students to corporate research labs can help solve the ML task (and, over the last 6 weeks, we have received over 300 emails from people who’d love to get involved).
- Help to put good solutions into production, once the community has developed them, so we can be reducing emissions ASAP.
Our first area of focus: Solar Photovoltaics
Solar Photovoltaics (PV) is the largest source of uncertainty in electricity system operators’ forecasts. If a dark cloud moves across the sky, the grid can be taken by surprise and lose hundreds of megawatts of PV generation within minutes. This lost PV generation must be replaced immediately. But thermal generators take hours to spin-up from cold. The end result is that, whenever the sun is shining, the grid keeps lots of spinning-reserve online: mostly gas turbines, which are kept idling, but not generating electricity. This is expensive and carbon intensive.
The grid would require less spinning reserve if they had better PV forecasts for the next few hours. That is, better PV forecasts would reduce carbon emissions, and save money. In the UK, better PV forecasts should save £1-10 million per year (Taylor et al, 2016), and about 100,000 tonnes of CO2 per year. Scaled up globally, the carbon savings should be of the order of tens of millions of tonnes per year.
Also check out Jack's blog post about getting involved.
Our intention is that Open Climate Fix will reduce emissions by at least millions of tonnes, at a cost of a few dollars per tonne, so considerably cheaper than most other interventions.
The RAAIS Foundation have very kindly given OCF our first grant!
“The RAAIS Foundation mission is to advance education and research in artificial intelligence for the common good.
We believe that the real-world impact of AI research will go well beyond solely for-profit applications.
The Foundation creates educational content for the general public to build awareness and knowledge about AI technology and its impact on the world.
The Foundation awards grants for open source research and projects that align with our mission. In particular, we support communities that would otherwise not have a chance to participate in advancing AI.”
If you are into climate philanthropy - or if you know of anyone else who is, please get in touch.
Jack Kelly is terrified by climate change and is determined to do everything in his power to reduce emissions, and so he has been applying machine learning to climate change mitigation for about ten years.
Jack left Google DeepMind in December 2018 to build Open Climate Fix. At DeepMind, Jack worked as a research engineer working on using neural networks to predict wind power.
Prior to DeepMind, Jack did a PhD and postdoc at Imperial College London on electricity disaggregation. During his academic work, Jack collected and released the UK’s largest dataset of domestic appliance-level electricity consumption, he was the first to apply deep learning to disaggregation, and he was co-founder and one of three lead developers on the most popular open-source framework for electricity disaggregation.
Dan Travers has moved to part time in his role as Senior Vice President of Product Management in financial markets software in order to work on solar energy forecasting in which he is currently a part-time PhD student at University of Sheffield.
Dan has 20 years of experience working in data analysis and analytic software engineering for energy and derivative markets trading and risk management, and now has a passion for using his vast experience to help to mitigate climate change.
Dan previously worked in an electricity trading and supply firm in Australia and has a Masters in Mathematics.