- Buildings
- Smart meter disaggregation
- Heating controls
- Modelling the energy usage of buildings
- Use robots to apply insulation to old buildings
- Demand response
- Football pitches
- Analyse and present data on the behaviour of appliances and machines in large buildings
- Building management systems
- Identify losses in district heating networks
- Transport
- Making computer systems more efficient
- Energy production
- Energy storage
- Energy system modelling
- Optimising the electricity grid (optimal power flow; unit commitment; economic dispatch; etc.)
- Agriculture
- Detecting environmental change
- Climate science
- Biology
- Computational materials science
This page is a non-exhaustive list of interesting subject areas for folks who, like me, are interested in both computer science and environmental sustainability. It aims to provide an (incomplete) answer to the question: How can a computer scientist contribute measurable improvements to our environmental sustainability?
Good overviews are provided by:
-
How Green is Your IT? in Computing Now, 2013
-
Bret Victor’s 2015 blog post on What can a technologist do about climate change?
-
Also see Zico Kolter’s 2012 video lectures on "Computational Methods in Sustainability"
Below is a list of topics:
Estimate an itemised energy bill from a single smart meter. Not clear how much energy this would save across the entire population though.
"Intelligent" controls which try to only heat the rooms which are occupied; and only do so when the house is occupied. For example, see:
Or just a concerted effort to make 'standard' heating controls more user-friendly would probably go a long way towards helping. How about an open-source, well designed controller that makes it really easy to understand what’s going on.
Demand response doesn’t reduce total energy consumption, but it can reduce carbon emissions by allowing the grid to accommodate more intermittent, renewable generation.
Football pitches heat the grass to optimise the growth rate. Existing pitches have very primitive controls. NEC presented work on this topic at BuildSys2015.
Hence allowing building managers to reduce energy consumption.
In the USA in 2012, commercial buildings used 2 trillion kWh of total energy (according to the EIA). In the UK, commercial buildings are responsible for 10% of the UK’s emissions (ref).
Companies already analysing the energy behaviour of large buildings include:
The basic idea is that buildings waste lots of energy (e.g. rooms which are simultaneously cooled and heated). And one might hope that commercial operators would be 'economically rational' and hence would be willing to invest a little time and money to save energy.
However, there is evidence that even commercial building operators aren’t very highly motivated to save energy. e.g. see the 2014 article in the Guardian on "Too big to save: why commercial buildings resist energy efficiency".
Large buildings tend to have building management systems. Making these more "energy aware" might help to reduce energy consumption.
e.g. this article: "UPS aims to safely deliver logistics clean tech through the 'Valley of Death'", Business Green, April 2016
e.g. tinkering with the Linux kernel scheduler; although it appears that the Linux scheduler has been very good at minimising energy usage since at least 2.6.38.
-
"Crayon: Saving Power through Shape and Color Approximation on Next-Generation Displays" by Stanley-Marbell, Estellers and Rinard, 2016.
-
Make compilers optimise for energy efficiency.
-
"Approximate Computation With Outlier Detection in Topaz" by Achour and Rinard, 2015.
IBM have done work on this: Dr Hamann from IBM spoke about this at SenSys 2015.
See:
-
Perera, Aung & Lee Woon 2014 "Machine Learning Techniques for Supporting Renewable Energy Generation and Integration: A Survey"
Doing a better job of predicting the output of renewable generation would help the grid to better utilise that renewable generation.
e.g. using video camera with a fish-eye lens looking up at the sky to predict when cloud cover will pass over a solar array - even being able to predict power output a few minutes into the future can help the grid to respond - Dr Hamann from IBM spoke about this at SenSys 2015.
See:
e.g. Yalmaz & Özer 2009.
e.g. see Review of computational fluid dynamics for wind turbine wake aerodynamics by Sanderse; van der Pijl & Koren; 2011.
e.g. Fengqi et al 2009.
e.g. see Growing Energy Labs Inc., based in San Francisco: "the same underlying technology that goes into modeling, designing and implementing storage projects will run the systems when they’re turned on, both to manage their technical performance and to track their financial performance."
Countries like India, where the grid is intermittent, are an interesting use-case for grid-scale battery storage. e.g. see "AES Energy Storage and Panasonic Target India for Grid Batteries" (20th April 2016) in Green Tech Media.
e.g. see the overview of the UCL Energy Institute’s Energy Systems team.
Scheduling generators to meet demand is tricky; especially given increasing levels of renewable generation on grids around the world. A fast & robust method for solving alternating current optimal power flow could save "tens of billions of dollars" according to this excellent introduction to the field: History of Optimal Power Flow and Formulations by Cain, O’Neill & Castillo, FERC, 2012
Also see Zico Kolter’s 2012 video lectures on "Computational Methods in Sustainability".
IBM have done work on using aerial photography to detect dry ground in a vineyard and then control water distribution to target the driest parts of the vineyard: Dr Hamann from IBM spoke about this at SenSys 2015.
Google Earth Engine is "a planetary-scale platform for Earth science & data analysis - Google Earth Engine combines a multi-petabyte catalog of satellite imagery and geospatial datasets with planetary-scale analysis capabilities and makes it available for scientists, researchers, and developers to detect changes, map trends, and quantify differences on the Earth’s surface." One use-case was detecting deforestation in "near real time".
Analysing today’s climate and forecasting future climates is very, very compute-intensive.
If we could speed up the design of new materials then that could help build better low-carbon generators and improve energy efficiency. See:
-
Nosengo, Can artificial intelligence create the next wonder material?, Nature, 2016
-
MACHINE LEARNING IN MATERIALS SCIENCE: RECENT PROGRESS AND EMERGING APPLICATIONS, 2016
-
MoleculeNet, a dataset which suggests AI may be ready to rapidly analyze molecules, learn their features, and classify and synthesize new ones
-
Citrine.io - "A.I. Powered Materials Informatics Accelerating The Global 1000"
-
“Deep Learning for Analysis of Materials Science Data” - “This project seeks to build automated, streaming analysis pipelines for extracting scientifically-meaningful insights from datasets relevant to materials discovery, especially x-ray scattering images. Our strategy is to develop data-analysis pipelines that serve a dual role: providing experimenters with useful (physically-meaningful) intermediate results, and using these analysis results as inputs to machine-learning methods.”
-
Automatic chemical design using a data-driven continuous representation of molecules - 2017 - They use unsupervised learning as “this lets us take advantage of large chemical databases containing millions of molecules, even when many properties are unknown for most compounds”. They use VAE. “The autoencoder was trained on a dataset with approximately 250,000 drug-like commercially available molecules extracted at random from the ZINC database. We also tested this approach on approximately 100,000 OLED molecules generated combinatorially.”
-
Machine Learning Method Accurately Predicts Metallic Defects - Feb 2017.
-
Solving the quantum many-body problem with artificial neural networks - Feb 2017
Books
-
Computational Materials Science: An Introduction, Dec 2016, 2nd Edition.
Courses
-
Coursera, Materials Data Sciences and Informatics
-
Coursera, “Material Processing”, “you will learn how a material’s properties are determined by the microstructure of the material, which is in turn determined by composition and the processing that the material has undergone.”