Renewable production forecasts “are essential for the balance of the electricity network” – Le Monde de l’Energie

Le Monde de l’Énergie opens its columns to Morgane Barthod, founder of meteo*swift, a start-up dedicated to forecasting solar production and wind turbinerecently acquired by Trailstone, to discuss with her this crucial issue of energy transition.

The World of Energy —On what mechanisms are the production forecasting techniques of intermittent renewable energies currently based? What are the latest breakthrough innovations in this area?

Morgane BarthodMy expertise focuses on solar and wind forecasts. Other sources of intermittent renewable energy include run-of-river hydroelectric plants. Forecasting techniques for these power plants involve modeling water inflow using rainfall forecasts, watershed modeling and sometimes sensors placed in watercourses.

For wind energy, we use both weather forecasts (from numerical weather prediction models or sensors), real-time data transmitted by wind farms, and artificial intelligence algorithms (family of timeseries ). We can thus model the behavior of each wind turbine in a given weather situation: the algorithm learns its past behavior thanks to the time series of weather forecasts and production, and deduces a forecast for the next day.

Other techniques are also used to improve the accuracy of forecasts, for example downscaling by computational fluid dynamics modeling to obtain finer mesh weather forecasts. An innovation in this area is the use of deep learning to take into account a broader weather spectrum: rather than taking weather forecasts from a single source point, algorithms can store forecasts for an entire area and thus capture weak signals (for example a weather disturbance which was to pass 10km north of the wind farm and which will finally pass over it can still be captured and taken into account by the models).

The main breakthrough innovation was the use of artificial intelligence in this area, an approach based on historical data rather than explicit modeling based on the wind turbine’s data sheet.

For solar forecasts, artificial intelligence is also used to model the behavior of photovoltaic parks. For intraday forecasts, we complete with cloud cover satellite data: we try to predict the evolution of existing clouds and anticipate if and/or when they will cast shadow on the solar panels.

The World of Energy —Why are weather forecasts crucial for predicting renewable energy production?

Morgane BarthodThe production of wind and solar energy depends directly on the weather: mainly the wind and the sun, but also, for example, temperature and humidity (indeed, if it freezes, frost or even a layer of ice can form on wind turbines, and slow their rotation or force operators to stop them to avoid the risk of projection), and other factors depending on the geographical areas. In some regions, for example, solar parks are heavily impacted by sandstorms since the sand covers the panels and prevents them from working.

It is therefore thanks to weather forecasts that we can calculate forecasts of solar and wind energy production.

Several horizons can be distinguished. Several days in advance or the day before for the next day, we will rather rely on numerical weather prediction models. A few hours in advance, we will also use satellite images for solar forecasting, which make it possible to identify where the clouds are and then apply a simulation model to anticipate their movements. A few minutes in advance, we will mainly use the real-time data transmitted by the solar and wind farms, to which we can add images taken by “fish-eye” cameras placed on the ground near the solar panels.

The World of Energy —What impact do these forecasts have on energy policies?

Morgane BarthodThese forecasts are essential for the balance of the electricity network and the large-scale integration of renewable energies into the electricity mix.

When wind or solar production undergoes strong variations but we know it in advance thanks to good forecasts, the electricity network can easily manage them in a low-cost and low-polluting way: by importing/exporting electricity, by storing/destocking (for example through pumped/turbine power plants), by modulating other energy sources (nuclear for example), by negotiating with factories for a temporary reduction in their consumption, etc.

When these variations are badly foreseen, there are only “emergency” solutions, which are more costly and polluting, for example gas-fired power stations or coal extra.

Forecasting is therefore a crucial link in the large-scale energy transition, and in the independence of the electricity grid from fossil fuels. The more the forecasts improve, the more it facilitates the integration of a large quantity of renewable energies into the electricity mix.

The World of Energy —What is the cost of advanced forecasting techniques, and is their adoption profitable for EnRi plants in relation to the savings they allow to be made?

Morgane BarthodThe cost is very variable depending on the quality of the precision (for very detailed forecasts, analysts and researchers can look at the data of each park and develop and implement improvements specific to each park), the number of parks planned at the same time (because then aggregate forecasts are then made, and the data from certain parks can be used to improve the forecast for other parks in the same portfolio), etc. However, these forecasts are very profitable and make it possible to make very substantial savings on the electricity markets (where forecasting errors are penalized by the network operator), which largely justify the investment in good quality forecasting.

We would like to thank the author of this article for this outstanding content

Renewable production forecasts “are essential for the balance of the electricity network” – Le Monde de l’Energie


Check out our social media accounts along with other related pageshttps://www.ai-magazine.com/related-pages/