The electricity system will be better able to manage local changes in supply from solar arrays as they react to moving weather systems, if a new solar ‘nowcasting’ service for the national control room is successful.
National Grid Electricity System Operator (NGESO) has teamed up with Open Climate Fix (OCF) – a non-profit start-up co-founded by former DeepMind researcher Jack Kelly – to use machine learning to improve its forecast of solar generation.
Changes in solar generation are difficult for grid operators to anticipate because of uncertainty both in forecasts and the location of solar panels, most of which are connected to regional networks. Work is under way to map the country’s solar panels, but there has been no way to anticipate short term swings in solar generation caused by cloud cover.

The ESO has to balance the electricity system second by second and has already used machine learning to predict rainfall hours and minutes ahead. OCF applies a similar approach to predict where sunlight will fall, by training a machine learning model to read satellite images and understand how and where clouds are moving in relation to solar arrays below.

With that knowledge the ESO is able to better manage the power kept in reserve – often gas plants – to respond to unexpected changes in supply or demand.

Carolina Tortora, head of innovation strategy and digital transformation at National Grid ESO, said: “Accurate forecasts for weather-dependent generation like solar and wind are vital for us in operating a low carbon electricity system. The more confidence we have in our forecasts, the less we’ll have to cover for uncertainty by keeping traditional, more controllable fossil fuel plants ticking over.
“We’re increasingly using machine-learning to boost our control room’s forecasts, and this latest nowcasting project with Open Climate Fix – whose work could have real impact for grid operators around the world – will bring another significant step forward in our capability and on our path to a zero carbon grid.”

Machine learning on reserve
The solar initiative follows another machine learning project in which NGESO is collaborating with the Smith Institute to forecast day-ahead reserve requirements.
Currently ESO sets reserve levels based on historical generation and forecasting errors, and adjusted by forecast renewable generation output. Dynamic Reserve Setting (DRS) can improve that by linking generation and forecasting to predictor variables, such as temperature and wind forecast data, to create more accurate predictions.
The model will help ESO better understand where there are uncertainties in its forecasting data, and set reserve levels more accurately – potentially limiting the need to keep fossil fuel plants running as back-up, reducing emissions and saving costs.
If the initial proof-of-concept proves successful, the algorithm will be integrated into ESO’s systems for a live trial.
The project is set to last for approximately twelve months, with the initial proof-of-concept expected in November 2021.