The UK鈥檚 forecasts for solar power generation have become far more accurate through the use of artificial intelligence, in a development that could lower energy bills and carbon emissions.
The country鈥檚 energy system is becoming more reliant sources of electricity with a variable output. Renewables like wind and solar, which depend on the weather, provided 36 per cent of our electricity at the start of this year, up from 7 per cent in 2009. 鈥淭he growth in solar was much, much more fast-paced than anyone anticipated,鈥� says Cian McLeavey-Reville at National Grid Electricity System Operator (ESO), which balances electricity supply and demand in England, Scotland and Wales.
Solar panels are connected to local distribution networks rather than the company鈥檚 national transmission network, meaning it has patchy visibility on solar. When solar panels are feeding in more electricity, that appears to the firm as reduced demand on the transmission network. Combined with the trickiness of forecasting the weather, that makes forecasting demand 鈥渕uch more difficult and volatile鈥�, says McLeavey-Reville.
Advertisement
But solar power forecasts have now been made a third more accurate, thanks to a partnership with the Alan Turing Institute. Previously, forecasts were based on just two pieces of data: installed solar power capacity and the amount of the sun鈥檚 energy that hits the Earth. 鈥淲hile this simple model can work relatively well for forecasts at very short times ahead, the accuracy degenerated quite rapidly,鈥� says Andrew Duncan at the institute.
Duncan helped build a machine learning model that trained itself on scores of variables, including temperature, historical irradiation data, location, and technical specifications of the solar panels. These were fed into an algorithm to produce forecasts of solar power entering the grid. Forecasts for seven days鈥� time were about 10 per cent more accurate straightaway, and 33 per cent more accurate when combined with further modelling.
This matters not just for National Grid. Balancing sudden changes in demand is expensive as reserve power stations suddenly need to be called on. Such balancing costs up to 拢1 billion a year, a price that is ultimately paid for through household energy bills. Better forecasting could reduce that by as much as 拢50 million. It should also lower carbon emissions, since most reserve power plants burn fossil fuels.
Solar power .聽But the current heatwave is not great for solar, because the efficiency of solar panels decreases as temperatures rise.
