Energisme, Eficia or even Metron. A handful of French players market machine learning platforms designed to optimize the energy consumption of tertiary and industrial buildings.
“The key to getting through the winter is general mobilization,” insists the minister responsible for energy transition, Agnès Pannier-Runacher, after the defense council on September 2. No fewer than 32 nuclear reactors are shut down due to corrosion or maintenance. Faced with this unprecedented situation, Xavier Piechaczyk, president of the electricity transmission network (RTE), estimates in Les Echos that a drop in consumption of around 15% is necessary this winter, at the most tense hours, in order to avoid power failure. On the government side, we are therefore betting for the moment on a chosen energy sobriety strategy. In her opening speech at the Meeting of French Entrepreneurs of the Medef at the end of July, Elisabeth Borne declared that France should “prepare” for measures of “rationing” of energy, specifying that “companies would be the first affected” . Faced with an unprecedented energy crisis, they could find a lifeline in AI. And this, both with a view to reducing their electricity consumption, but also with a view to anticipating possible cuts.
Among the French publishers of energy AI, Eficia is developing a cloud mode platform designed for the tertiary sector. Upstream, it federates temperature, humidity and luminosity indicators from Lora sensors installed on all of its customers’ buildings. At the same time, it retrieves the energy consumption levels of the heating-ventilation-air conditioning (HVAC) and lighting systems, etc. All in real time. “If these devices are not communicating, we will equip them with automatons to control them remotely”, specifies Alric Marc, president and founder of Eficia. Behind the scenes, the learning model takes advantage of the history of the building, but also of the behavior of equivalent buildings of other customers, who will have been confronted with identical telemetry contexts. Eficia manages some 3,000 buildings to date, across France, Spain and Italy.
The peak activity challenge
Over time, Eficia’s machine learning algorithms get sharper. “The models estimate the inertia capacity of the building better and better, the objective being to regulate, or even cut off, the heating or the air conditioning with regard to the time necessary for each zone (room, floor, technical room, editor’s note) reaches the desired temperature, up or down, depending on the weather forecast”, explains Alric Marc. “The heat is rising. As a result, it will not be necessary to heat the ground floor as much as the upper floors.” A criterion that the AI will integrate into its equation. To this is added another dimension: activity. the CO2 concentration will make it possible to estimate the number of people present per zone according to the time of day, the day of the week, and the period: holidays, public holidays, telework slots. adjust the HVAC system (heating, ventilation, air conditioning) By noting potential discrepancies in the readings, the AI will also identify a window left open or a potential technical problem: faulty circuit breaker, light left on at night, HVAC failure , etc.
“In the case of a campus, AI will be able to predict consumption/production patterns between buildings”
Faced with Eficia, Metron claims a much more generalist positioning. Alongside the tertiary sector, the publisher, also Parisian, also targets industrial infrastructures. The challenge: to optimize the energy consumption of manufacturing processes while maintaining the highest product quality. “At Arcelor, for example, we help operators adjust the opening time of ovens. A few seconds more or less can have a substantial impact on electricity expenditure. Knowing that many parameters must be taken into account in parallel that can vary this adjustment: the quantity of molten fluid in the furnace, the humidity level in the air, etc.”, explains Tanguy Detroz, CEO of Metron. “The logic is equivalent for pasteurization processes in the food industry. Ditto for the adjustment of the valves of wastewater treatment plants which must take into account multiple factors: rainfall, weather, volume, oxygen level in the sludge. Driving 20,000 sites on four continents, Metron’s AI combines several types of algorithm: decision tree, linear regression and neural network.
Manage load shedding in the event of a shortage
Another French player to put AI at the service of energy savings, Energisme, like Metron, targets the tertiary sector and industry. Based in Boulogne-Billancourt, this company even manages electricity production systems. “Our SaaS N’gage offer takes into account the capacities of the buildings’ photovoltaic panels, and those of the on-site electric car charging stations. Terminals that can be bidirectional and use vehicle batteries to power the building”, emphasizes Thierry Chambon, Managing Director of Energisme. What level of sunshine is predicted by the weather forecast? How many electric cars are there and how charged are they? Based on these indicators, N’gage can choose one current source over another. “In the case of a campus, the AI will be able to predict consumption/production patterns between buildings. If it anticipates that one of them will have few or no employees over a given period, due to teleworking for example, it will then route its electricity production to others, depending on their level of activity”, adds Thierry Chambon.
At Eficia, we promise average energy savings of 20% to 25%, with the key to a return on investment after 6 to 48 months. “The price of the project will depend on the perimeter and above all on the complexity of the buildings to be covered”, comments Alric Marc. It remains to be seen how to respond to potential future rationing measures that would be taken by the government this winter. “We are currently working with our customers on load shedding plans, aimed at switching off or reducing the consumption of certain non-vital equipment, over given time slots, in order to relieve the electricity network in the event of a shortage”, confides Alric Mark. Eficia’s AI will ingest this constraint as a new parameter to take into account.