Predict the weather can be a frustratingly imprecise science. The weather app on your phone is good enough for predicting if it’s likely to rain at any given time during a given day, but much less useful if you want to know if there will be a downpour in the center of London at 3 p.m. this Sunday. If you absolutely have to stay dry, you might be better off keeping an umbrella with you or staying indoors.
For most people, not knowing what the weather is going to do in the next hour is a minor inconvenience. But when it comes to the power grid, not knowing what the weather will do next isn’t just a nuisance, it’s a major source of carbon emissions. If we could better predict when and where the weather will change, we could prevent huge amounts of carbon dioxide from being released into the atmosphere simply because we don’t know what the clouds are going to do next.
Here is the problem. On a sunny spring day in Britain, solar power can account for around 30% of all electricity produced on the island. The exact number varies a lot, but under ideal conditions – solar panels perform best on cool but sunny days – they can produce 9 gigawatts (GW) of power, a huge chunk of the average power demand of 30 GW. So far, so good. But if a large cloud descends over the southwest, where Britain’s many solar panels are located, a significant portion of that renewable energy suddenly disappears from the grid – the equivalent of instantly taking a gas-fired power station. Hundreds of megawatts of energy are gone like this.
Losing power to an entire power plant in a matter of minutes is obviously not ideal, so to compensate for this, power grids schedule back-up power generation to step in and alleviate the bumps caused by changes in the power supply. solar production. In Britain, the responsibility for balancing and distributing this energy lies with the National Grid Electricity System Operator (ESO), which requires fossil-fuel power plants, usually natural gas, to produce additional energy by case of an unexpected drop in solar production.
Fossil fuel plants are slow beasts. “We would really like to have a power plant that can ramp up in five minutes or half an hour, because that’s how quickly wind and solar power generation can change,” says Jan Kleissl, professor of energy. renewable and environmental flow at University of California at San Diego. But fossil fuel plants don’t work that way. They take a long time to light up and are most efficient when operating at full power. This limitation further encourages power grids to overproduce power just in case solar or wind power drops.
One way to get around this problem is to forecast the weather better. If we knew exactly how much solar power Britain was likely to produce at any given time, ESO could reduce the amount of energy it holds in reserve, thereby reducing the total carbon footprint of the power grid. In other words, if we knew exactly how much solar energy was going to flow through the grid every five minutes, we could make sure we were using every kilowatt of that energy rather than hedging our bets with the excess electricity generated. by fossil fuel power plants.
Jack Kelly thinks he knows a way to dramatically improve those predictions. A former researcher at DeepMind, the artificial intelligence company owned by Alphabet, in 2019, Kelly co-founded Fix for open climate, a non-profit organization focused on reducing greenhouse gas emissions through machine learning. “I’m a machine learning researcher who is terrified of climate change and eager to do everything possible to try to fix it,” Kelly says. He believes that better solar forecasts in the UK could save 100,000 tons of carbon dioxide emissions each year, and will be critical if the National Grid ESO will meet its 2025 goal of operating at zero emissions whenever there is enough renewable production available.