Sustainability is a significant challenge and touches a broad range of domains.
By Dr Alex Kell
Sustainability is a significant challenge and touches a broad range of domains. Some could be the limiting of the effects of global warming by reducing electricity consumption, switching to renewable energy, or preserving biodiversity. It is true, therefore, that a single approach won’t solve this broad range of issues. However, it may be possible to use generic computer science methodologies to tackle these issues from various angles, particularly with big data analytics. This is what the first “Big Data Analytics for Sustainability workshop” at the IEEE Big Data 2021 conference set out to tackle. The workshop was organised by Dr Alex Kell (Sustainable Gas Institute, Imperial College London), Dr Matthew Forshaw (School of Computing, Newcastle University and Dr Stephen McGough (School of Computing, Newcastle University).
The workshop began with a keynote presentation from Diego Moya et al. from the Sustainable Gas Institute, Imperial College London. Diego presented his paper on the use of geospatial big data analytics to model the long-term sustainable transition of residential heating worldwide. This paper used an innovative approach to assess global energy demands spatially and temporally in the residential sector, with a particular emphasis on space heating. This high-quality data was then used within the MUSE energy system model.
Another important aspect tackled in this workshop was the automatic forecasting of wind speeds for wind turbines. Rathore et al. from TCS Research, India presented a paper titled “Multi Scale Graph Wavenet for Wind Speed Forecasting”. In this work Rathore et al. propose a novel deep learning architecture to predict wind speeds at various locations in Denmark based on both spatial and temporal relationships. They found that by considering dependencies between both regions and time they were able to outperform state-of-the-art methods on wind speed forecasting for multiple forecast horizons by 4-5%.
Moving away from the energy theme, Curry et al. used deep learning to identify the Scottish wildcat in the wild. A major issue with this approach, however, is that Scottish wildcats are so rare that there do not exist many images of them to train the machine learning algorithms. Curry et al. therefore propose the generalisation of images taken in captivity to identify the animals in the wild. However, they conclude that models trained on captivity imagery cannot be generalised to wild camera trap imagery using existing techniques. However, they identified a possibility for further research to be carried out to improve accuracies.
The papers presented here are just a few of the papers presented, with other solutions proposed for the automatic control of boilers and the forecasting of pollution. The ability for big data to tackle problems related to sustainability were demonstrated at this workshop, with further advances critical to the development of a sustainable society. For more information and links to the papers refer to the workshop’s website.
Article text (excluding photos or graphics) © Imperial College London.
Photos and graphics subject to third party copyright used with permission or © Imperial College London.
Department of Earth Science & Engineering