This text is part of the special Energies notebook
Quebec supports research to help the French-speaking world move towards renewable energies.
Artificial intelligence (AI) can help us improve efficiency across many parts of society. Particularly in the management of renewable energy, essential in the context of climate change. To achieve this, several research projects are being carried out.
While it is still difficult and expensive to store the electricity produced for consumption later, the ideal is to align production with needs. To achieve this, you still need to be able to predict electricity demand. This is what Ayoub Atanane is working on as part of his master’s degree at the University of Quebec at Rimouski (UQAR), under the supervision of Loubna Benabbou, professor in the Department of Management Sciences.
His master’s project is one of those funded by the International Climate Cooperation Program (ICCP) of the Quebec Ministry of the Environment and the Fight against Climate Change to strengthen the capacity of French-speaking countries to adapt to the effects of climate change. No less than $840,000 was granted in 2020 to Mila, IVADO, Polytechnique Montréal and UQAR so that researchers can contribute to a broader deployment of renewable energies and support Morocco in its efforts to reduction of its greenhouse gas emissions.
To train his forecasting model using AI, Ayoub Atanane needed open access data of different types that influence electricity demand. We think, for example, of the history of electricity consumption, weather conditions, demographics, the calendar of major events, such as sports matches, etc. He found them in different countries, such as the United States, Australia and Turkey. Result ? “We have proven that the model works well,” he notes.
Now that the model is trained, it could be used by different administrations. “All you have to do is enter the data for a location, whether in Morocco or elsewhere, and the model will be able to predict electricity demand,” he says.
Forecast photovoltaic energy production
In the same PCCI-funded program, Saad Benslimane looked at deep learning forecasting of solar energy production. It was as part of her master’s degree carried out under the supervision of Hanane Dagdougui, professor in the Department of Mathematics and Industrial Engineering at Polytechnique Montréal.
“The goal was to create an open-access tool that can predict photovoltaic energy production,” explains the man who now works as a research and technology scientist at InnovLOG, which helps companies innovate to improve their logistics.
He adds that the advantage of having access to good forecasts of solar energy production is that the manager can better plan production, see when he will be able to sell energy and also plan the maintenance of his equipment.
To create this model, Saad Benslimane found quality, freely accessible data in Australia, particularly on wind speed and direction, and then on atmospheric pressure. The data also included weather conditions one hour before starting the forecast, as well as the panels’ electricity production history. When he looks at the results, he sees that the forecast horizon greatly influences the accuracy.
“If we make predictions for the next 15 minutes, the model generates a value per minute,” explains Saad Benslimane. Accuracy decreases as the prediction horizon increases, for example if the horizon is 24 hours or one week. »
The researcher assesses that this weakness is caused by the fact that the model does not have data on weather forecasts, whereas if, for example, the afternoon looks cloudy, the capacity to produce energy solar will be less.
“To have a more precise model, it would be interesting to integrate satellite weather forecast data,” indicates Saad Benslimane.
Then, to move towards the adoption of this type of model in the industry, he sees a big obstacle. “Often, managers are used to working with software and they don’t want to change the way they do things,” he notes. We need to convince people to start using AI so that they can be more effective in carrying out their tasks. »
This content was produced by the Special Publications team at Duty, relating to marketing. The writing of the Duty did not take part.