Forecasts of bad weather are a disaster for the climate crisis

Weather forecast science can be inaccurate. Your phone’s weather app is pretty good at predicting that it may rain at some point on a given day, but it’s much more helpful if you want to know if there will be a downpour in central London this Sunday at 3pm. If you need to stay completely dry, you’d better have an umbrella with you or be inside.

For most people, not knowing what the weather will do next hour is a small inconvenience. As for the power grid, not knowing what the weather will be like next time is not annoying, it is a significant source of carbon emissions. If we had better anticipate when and where the weather would change, we could stop sending huge amounts of carbon dioxide into the atmosphere simply because we don’t know what the clouds will do.

Here is the problem. In Britain on a sunny spring day solar energy can generate about 30% of all electricity generated on the island. The exact amount varies a lot, but under perfect conditions — solar panels are best on cool but sunny days — they can extract 9 gigawatts (GW) of energy, a large portion of the average 30 GW of energy demand. So far so good. But if a large cloud falls to the south-west, where there are many solar panels in Britain, a significant part of that renewable energy suddenly disappears from the grid – the equivalent of the entire gas station that immediately goes offline. Hundreds of megawatts of energy were lost, just like that.

Eliminating the power of the entire plant in minutes is very convenient, of course, to compensate for this, the power grids organize the production of energy in the amount of safety to smooth out the shocks caused by changes in solar production. In the UK, the responsibility for balancing and distributing this energy lies with the National Electricity System Operating System (ESO), a fossil fuel power plant that typically burns natural gas, requiring additional energy to produce if solar production falls sharply.

Fossil fuel plants are slow-moving beasts. “We’d really like to have a power plant that can be expanded in five minutes or half an hour, because it can be a fast speed to generate wind and solar energy,” says Jan Kleissl, a professor of renewable energy and environmental flows at the University of San Diego, California. Fossil fuel plants don’t work that way. They take a long time to activate and are most effective when running at full power. This limitation encourages the grid to produce too much energy if it is blown away by solar or wind energy.

One way to achieve this is to forecast the weather better. If we knew that Britain would be able to produce solar energy at any time, it could reverse the power that ESO has in reserve, lowering the entire carbon footprint of the energy grid. In other words, if we knew exactly what solar energy would be emitted into the grid every five minutes, we could be sure that we would use every kilowatt of that energy instead of covering our bets with excess electricity generated from fossil fuel power plants.

Jack Kelly believes he knows how to improve these predictions a lot. DeepMind, a former researcher at the Alphabet-owned artificial intelligence company, co-founded Kelly in 2019 Open Climate Fix, using automated non-profit learning aimed at reducing greenhouse gas emissions. “I’m a researcher learning machine that is terrified of climate change, and I want to do everything I can to fix it,” Kelly says. He estimates that better solar forecasts can be made in the UK save 100,000 tons since it emits carbon dioxide every year, and will be critical if the National Grid ESO meets its target of zero emissions by 2025 when it is sufficient. renewable creation available.