Water: a major challenge for cities
The destiny of civilisations would seem to be closely entwined with their ability to manage water. The Roman aqueducts, Arabic gardens (such as those in the Alhambra with their ingenious irrigation systems) and even the Parisian sewers (designed by Baron Haussmann) are each an illustration of the greatness of their day. But before being a source of pleasure, power or wealth, water is a source of life. And yet, it is still largely improperly managed. Even worse: each year, 2.2 million children die due to a lack of access to healthy water.
This commodity – both rare and abundant – is one of the main challenges cities are facing. So, how can AI help us improve the way water is managed? To start to answer this question, I visited the CANN Forecast startup in Montreal (Canada) and the Baumette wastewater treatment plant in Angers (France). The latter facility is notable for its ability to transform wastewater into valuable resources (electricity and fertiliser) for the local region.
From predictive maintenance…
The Baumette wastewater treatment plant has been using AI for the predictive maintenance of its facilities for almost a year. Operated by Veolia, the plant uses connected boxes, called BOB, to anticipate mechanical failures. BOB was developed by the Cartesiam startup and Eolane. This technology uses internal AI (NanoEdge). Before becoming operational, the solution must undergo a period of unsupervised learning. For almost seven days, BOB listens to the vibration signal from the piece of equipment it will be monitoring. At the end of the learning phase, the connected box can report any discrepancies that could result in machine failure or malfunction.
In addition to being particularly easy to use (BOB just plugs into the machine), it consumes very little energy, with one battery lasting three years. The pre-emption period set for BOB reflects the extent of the environmental or economic impact that equipment failure can have on a plant. For example, the failure of an air compressor may affect the entire plant’s process and pollute the Maine, the river where the plant’s treated water normally flows.
CANN Forecast uses a different branch of AI (supervised learning) for the predictive maintenance of water mains. “In Montreal, as in most large cities, drinking water pipes are reaching their end‑of-life. When a pipe ruptures, the city loses resources and money,” said Naysan Saran, co‑founder of CANN Forecast. In 2018, the cost of repairing water mains in North America added up to $3 billion. To mitigate this problem, CANN Forecast developed a proactive solution called InfoBris, which is five times more accurate than conventional methods. The startup, which already works with around ten cities (including Montreal) on this issue, can provide its partners with a list of water mains and their rupture probability. Its AI system analyses past breakage data together with locations and environmental conditions. CANN Forecast created a huge ecosystem (including Mila, McGill University and Environment Canada) to design this solution.
…to smart water management
As well as InfoBris, CANN Forecast has developed InfoBaignade. This system can predict, with 95% accuracy, the concentration of E. coli, which can be used to determine whether river water is swimmable or not. “Currently, it takes between 18 and 48 hours to get an E. coli concentration reading. In a dynamic environment, such as a river, water contamination can change from one day to the next, so this method is inefficient,” said Nicolas St-Gelais, co-founder of CANN Forecast. By aggregating around ten different types of data (rainfall, overflow, and other variables), the Montreal startup is able to provide a smarter water management model that factors in the dynamic nature of rivers.
BOB is also used to optimise equipment management. The system can provide long‑distance monitoring of remote equipment and replace periodic inspections with occasional targeted servicing.
Augmented rather than artificial intelligence
It is interesting to note that BOB does not provide data but rather insights via a LoRa network. This means the box users have access to an easy-to-read and readily understandable platform. This ease of use is crucial to the solution’s proper operation – and that’s because,according to Joel Rubino, co-founder of Cartesiam, “BOB acts as an assistant to the plant employees.”
The same applies to CANN Forecast: “We are trying to understand how we can incorporate it [InfoBris] into cities’ decision-making processes,” said Nicolas St-Gelais. For both Cartesiam and CANN Forecast, the aim is to augment decision-makers’ knowledge and rather than replacing them