This aspect of artificial intelligence gives computers the ability to "learn" on their own from the data they are fed. It is used by cities to reduce their carbon footprint and improve resource management.
The brainchild of a workshop held at the International Machine Learning Conference in June 2019, the white paper entitled Tackling Climate Change with Machine Learning explains how artificial intelligence technology can help reduce greenhouse gas emissions in a variety of areas. In particular, technological advances help cities’ reduce their carbon footprint.
Machine learning is therefore an essential tool that helps building managers and policy makers to improve the energy efficiency of a group of buildings.
Urban building energy modeling, used in particular to manage district heating and cooling networks (as in the EU-supported InDeal project), means it is possible to analyze the energy performance of several buildings on scales ranging from a block or a neighborhood to an entire city. On the basis of various characteristics - building use, dimensions, materials, roof type, immediate environment, etc. - machine learning can help predict the energy consumption of the entire area under consideration.
For example, from the data released by commercial and residential building owners under the disclosure laws passed by several U.S. cities, researchers at the NYU Tandon School of Engineering used various machine learning methods to predict the energy consumption of 1.1 million buildings in New York City. The data collected can also be used by algorithms to identify buildings with the highest renovation potential.
Better urban data management
In addition, cities are increasingly using connected objects and new information technologies to improve the quality of urban services and manage resources more efficiently.
To do so, they are now collecting an ever-increasing amount of data from a variety of sources: sensors in different locations that measure, for example, road traffic or pollution levels, as well as mobile devices and applications used by city dwellers.
In the smart city, this data is used as a basis for developing urban policies covering mobility, energy, water, material flows and waste management - and makes it possible to create circular economy loops.
The volume and great diversity of the data, however, results in some challenges that machine learning is able to address through techniques such as pre-processing the data (and so transmit only what is relevant), improving knowledge extraction, and refining indicators and predictions. And all that while optimizing the energy consumption of the techniques themselves.
Optimizing waste management
In addition to the numerous examples cited in the International Conference on Machine Learning report, smart waste management - i.e. the use of digital and technological innovations, in particular artificial intelligence and machine learning to optimize waste management - also helps to reduce greenhouse gas emissions.
Essential for any smart city, good waste management allows local authorities and specialist companies to sort, collect and treat waste on the basis of a circular economy approach and at the same time reduce their costs.
The Max-AI smart sorting robot, tested by Veolia in several of its sorting centers in France, is a good illustration. The Group has developed several technologies, such as automatic sorting for packaging using material and color, and remotely controlled sorting to improve sorting performance and increase waste recovery rates.
Driven by machine learning, Max-AI, the "association of an "eye", a simple optical camera, and an "arm", an articulated robot controlled by a "brain", a neural network implanted in a computer", completes these solutions. In order for it to become progressively more autonomous and versatile, operators teach it to distinguish each type of waste by enriching its database with hundreds of thousands of images.
CREDITS: Main picture © Noémie Rosset / Veolia