What’s happening at Open Climate Fix? An Update.
There is certainly a lot going on in the world right now, and we hope everyone reading this blog post is in good health. While it is hard to think about anything else right now, the challenge of climate change is no less urgent. We wanted to give an update on what's happening at Open Climate Fix.
Firstly, it’s encouraging to see emissions going down in the short term, we are hoping to create a world where we can lock in those gains for the long term! Many things have happened here at Open Climate Fix since our last blog post, so here’s an update. Let’s look at the current status of our two big projects: Nowcasting and PV Mapping.
Let’s start with our two projects: They both directly work towards our current goal, which is to predict how much electricity will be produced by solar photovoltaic (PV) panels over the next few hours (called "nowcasting"). More about why we work on nowcasting can be found on our project page. In short: Solar Photovoltaics are awesome, but it is hard to predict their power output. A big cloud will cause solar electricity generation to drop. That’s bad, because electricity supply must match demand at every moment.
To protect against drops in PV electricity generation (read: big clouds), electricity grid operators instruct some fossil-fuel generators to run below their maximum capacity. If PV generation drops, grid operators instruct these "spinning reserve" generators ramp up quickly to cover the lost PV generation. Supply and demand are kept in balance.
While “spinning reserve” sounds great at first, that usually means having a fleet of natural-gas turbines running at low efficiency (gas turbines are less fuel-efficient and carbon-efficient when throttled back). If we knew with greater certainty that there won’t be any big clouds in the next few hours and thus PV generation won’t go down, we could tell the grid operators to commit to a smaller number of generators, each running close to their maximum capacity; rather than committing to a large number of generators running below their maximum capacity and hence running at low efficiency.
We estimate, conservatively, that better PV forecasts could reduce UK CO2 emissions by 100,000 tonnes CO2 per year (although this number is very hard to estimate with precision!). This estimate goes up to 100 million tonnes of CO2 globally by 2030. More about our assumptions and how we got to that estimate can be found in our “original” OCF doc.
To predict PV generation for any point in the UK we made the following progress:
- Acquired and parsed tens of terabytes of satellite imagery & PV data.
- Initial results from our own machine learning research indicates the approach is viable. More to come!
- Collaborated with amazing students on the Stanford AI for Climate Change Bootcamp where 3 Stanford students worked on solar PV Nowcasting for 6 months. Their results are very promising!
- Established with stakeholders that the nowcast fills a gap in current knowledge, and we will continue to interview potential users of nowcasting to better understand their needs (and this research will be published!)
- We have ongoing collaborations with two more awesome students!
In the near future, we hope to release a public dataset of satellite imagery & PV data, and run a public ML competition on Nowcasting.
To estimate how much photovoltaic energy will be produced, we first need to know where the PV systems are. At the moment this is not well known in the UK for all but the largest PV systems - particularly domestic systems are poorly documented.
Here is the progress we made with the PV mapping project:
- We supported a project on OpenStreetMap to map solar PV systems. Volunteers mapped 100,000 new PV systems in the UK in just 3-months, & continue to add PV systems to OSM. We were blown away! Huge thanks to Dan Stowell for leading this project!
- We recently finished a 3-month project with partners at the Alan Turing Institute, Sheffield University, and Queen Mary University which developed code to combine several public UK PV maps into a single map. The code is public and we plan to publish a 'merged' PV map soon.
- We supported a 2019 UCL MSc research project on using machine learning to detect solar domestic PV systems in aerial imagery.
As a next step we are looking into running a public ML competition on locating solar PV systems in aerial imagery soon, and to continue working on merging existing databases of solar PV.
One progress update we didn’t mention yet for Nowcasting is actually quite a big one: We got funding from the European Space Agency! As you might have seen on Twitter, the ESA has awarded us an AI Kick-Start co-funding for six months.
We’re over the moon that the @ESA has awarded us an AI Kick-Start co-funding for six months! 🚀— Open Climate Fix (@OpenClimateFix) April 6, 2020
Kick-Start activities are compact Feasability Studies to explore new service concepts that use space tech.
Our research will be public, of course! 😉 @ESAbusinessapps @Space4Europe
That is awesome news and will definitely help us accelerate working on Nowcasting. We will be reaching out to various partners and potential users of Nowcasting and interview them on their needs and wants.
Despite everything going on in the world right now, we are still positive and are continuing to work on fixing the climate. We talked to so many people over the last couple of months, open sourced seven different GitHub repositories and started Nowcasting and PV Mapping as projects. We already took a lot of steps and got awarded our first co-funding.
Make sure you subscribe to our newsletter if you want to stay up to date. We even have an RSS feed for our blog now, thanks to Grant (@grantbdev) on GitHub. We will be picking up blogging more actively again, so definitely stay tuned.