On 10th of June, 2019, twenty-two AI researchers, including Andrew Ng and Yoshua Bengio, published a paper on how to tackle climate change with machine learning. I really enjoyed reading it and I am convinced that the paper as well as the climatechange.ai initative, which emerged from it, deserve more attention. For that reason i created a series of blog posts and videos which provide a dense summary, listing many of the proposed solutions and linking research work as well as ongoing projects. In the big picture, all solutions aim to reduce greenhouse gas emissions.
As my contribution to the global #ClimateStrike week from September 20th to 27th, i will post one chapter (video and blog post) on every working day. You can subscribe to our YouTube channel or follow me on Twitter to be notified when new content is published.
This is part three of a six-part series:
- Electricity Systems
- Buildings & Cities
- Farms & Forests
- Industry & Carbon Dioxide Removal
- Datasets & further resources
Buildings & Cities
Buildings are responsible for about ¼ of global energy-related greenhouse gas emissions but there is huge potential in reducing emissions in existing and new buildings. However, local governments, who can shape cities and pass emissions reducing regulation, often need to act without sufficient data about their city. They could benefit from emission-relevant high quality data about their city and buildings. It would allow them to train algorithms and make smarter decisions.
Machine Learning can help reduce the carbon footprint of buildings and cities by
- ↓ modelling the energy consumption of buildings
- ↓ enabling smart buildings
- ↓ understanding the energy consumption of cities
- ↓ gathering infrastructure data
- ↓ collecting data for smart cities
- ↓ improving low-emission infrastructure
Modelling the energy consumption of buildings
Modelling the energy consumption of buildings help to evaluate building designs and operation strategies.
Machine Learning can help
- forecast energy demand for individual buildings from data produced by meters and home energy monitors, which is useful when evaluating building design and operation, but can also inform grid operators [Paper ]
- predict energy consumption of building concepts and generate more efficient building concepts, also transfer knowledge from commercial to residential buildings, from gas- to electricity-heated buildings [Paper – Reinforcement Learning, Deep Belief Networks, Transfer Learning]
Enabling smart buildings
The majority of energy consumed by buildings is caused by heating, ventilation and air conditioning (HVAC). Smart buildings try to reduce HVAC energy consumption.
ML can help
- forecast what temperatures are needed and thereby improve control. ML could also help detecting faults [Paper – Deep Belief Networks, Ensemble learning]
- achieve optimal control of hot water systems [Paper – Reinforcement Learning]
- improving efficiency of cooling systems by identifying faults such as refrigerant leakage [Paper – Bayesian Network]
- schedule efficient energy use based on data from sensors / IoT devices [Paper – Reinforcement Learning]
- adapt devices which consume energy to usage patterns, such as occupancy patterns [Paper – Decision Tree] [Paper – Decision Tree]; via Wi-Fi-enabled IoT devices [Paper – Autoencoder Long-term Recurrent Convolutional Network]
Potential pitfall: Smart sensors use energy, energy savings have to be higher than the consumption of energy consumed by all sensors. One has to be aware of security and privacy risks as well.
Understanding the energy consumption of cities
City planners could take better actions if they would be provided with more precise information about their city’s energy consumption, for example on district-level, ranking buildings based on their energy efficiency.
ML can help
- predict the energy use of a city’s buildings [Paper – Linear regression, Random Forest, Support Vector Machine], [Paper – Clustering, Supervised Learning], [Paper – Gradient Boosting Machine (XGBoost)]
Gathering infrastructure data
Classified buildings in french city [mapuce.orgbisgis.org]
There is no infrastructure data of many world regions. Data can be gathered using satellite images and radar data, which is available and relatively consistent for most regions.
ML can help
- predict energy consumption from features such as property class or presence of central air conditioning [Paper – Gaussian Process Regression]
- inform building energy retrofit policy makers by classifying buildings more accurately than conventional models by energy retrofit potential [Paper – Autoencoder, Neural Network, Clustering]
- to roughly classify all buildings worldwide [Paper ]
- segment exact building footprints at a national scale [Github-Repository – Convolutional Neural Network (ResNet34)]
- create 3D models of buildings from LiDAR data [Paper – Decision Tree, Random Forest, Support Vector Machine], [Paper – Support Vector Machine]
- Classification of french cities: http://mapuce.orbisgis.org/
- World urban database project: http://www.wudapt.org/
Collecting data for smart cities
For cities to evolve into smart cities and therefore understand emission related activities, they need to be able to access relevant data. As an example, the city of Los Angeles passed a regulation which requires vehicle-sharing companies such as scooter-sharing providers to provide data on location, use and condition of all vehicles via an open-source API [GitHub-Repo ].
ML can help
- improve urban policy making by helping cities deal with high volumes of data in real-time, for example with data stream classification [Paper – Ensemble Learning]
Improving low-emission infrastructure
In order to create low-emission infrastructure, which might for example include an optimal placement of electronic charging stations or optimal district heating, local governments can benefit from ML when conducting smart city projects. Smart city projects are mostly conducted in the wealthier global North, though the highest potential of climate change mitigation is in the global South.
ML can help
- transfer climate solutions across cities (from Global North to Global South) by clustering cities based on climate-relevant factors. [Paper – Clustering]
- improve public lighting systems by regulating light intensity based on historical patterns of foot traffic [Paper ]
Many thanks to all researchers of the paper:
David Rolnick, Andrew Y. Ng, Priya L. Donti, Lynn H. Kaack, Kelly Kochanski, Alexandre Lacoste, Kris Sankaran, Andrew Slavin Ross, Nikola Milojevic-Dupont, Natasha Jaques, Anna Waldman-Brown, Yoshua Bengio, Alexandra Luccioni, Tegan Maharaj, Evan D. Sherwin, S. Karthik Mukkavilli, Konrad P. Kording, Carla Gomes, Demis Hassabis, John C. Platt, Felix Creutzig and Jennifer Chayes.
Your job at codecentric?
More articles in this subject area
Discover exciting further topics and let the codecentric world inspire you.