Machine Learning and Artificial Intelligence for Intelligent Energy Systems

This project falls within the EPSRC digital economy and energy themes and addresses the application of machine learning and artificial intelligence within intelligent energy systems. The focus will be applications addressing the need for intelligent energy services within homes; predicting and optimising energy use and helping consumers change their energy consumption behaviour.

The project will seek to apply approaches such as Gaussian processes to predict future energy use within the home, exploring how additional information from sensors placed within the home, and that available from present and future smart meters, will help place such information in the context of consumer behaviour. In particular, exploring how the time resolution of smart meter data affects its usability in a range of applications; from 10 second data available through a local home network, through to 30 minute data available through the official DCC. It will explore how correlations between multiple time series within the home, and other public data sources, such as weather data or overall electricity grid demand, may be incorporated into predictions of energy use within any individual home. It will consider approaches to anonymity, such that learning may occur across homes, without sharing sensitive private energy consumption data.

In addition, it will explore the use of model-free approaches to understanding energy use, particularly energy used by heating systems, within domestic and commercial buildings. Current approaches typically assume detailed thermal models of the buildings based upon detailed construction data. However, for existing building, this level of detail is either impossible or too expense to obtain through manual survey. Thus, the project will explore machine learning approaches for model-free control of such buildings; learning a control policy through an exploratory learning period.

The project will involve close collaboration between the student at the University of Oxford and researchers at the EDF Energy research and development site in Brighton, UK. The collaboration will seek to ground the research conducted at the University of Oxford within the constraints of the real-world application, particularly that of an energy retailer operating within the UK energy market. The research will seek to deliver real solutions to existing challenges in this sector, and the student will work with researchers at EDF Energy to transition this research into business operations and real products. To facilitate this, the student will spend extended periods of time working alongside EDF Energy researchers at their research centre in Brighton. This will likely consist of one or more multi-week visit during the first year of the DPhil programme, increasing to internships of two months over the summers in the second and third year of the DPhil programme. In parallel, the University of Oxford will host short visits from EDF Energy researchers, allowing them to share their research results and to engage with other researchers working in similar areas.