DPhil in Energy Data Privacy
Supervisors: Philipp Grunewald, Zeynep Tekler, Tina Fawcett
Research topic
This inter-disciplinary doctoral studentship researches privacy of energy data with the aim to deliver better value for citizens, while safeguarding personal information and preventing data driven discrimination.
Over half of UK homes now have smart meters, which collect personal half-hourly gas and electricity data. These data contain a wealth of information, which can help households in making technological and behavioural choices. The right access to data, and appropriate means to communicate the relevant insights to households by way of feedback, is claimed to have the potential to support households on their journey towards zero-carbon living.
However, privacy concerns and regulatory arrangements have made access to these data difficult for private individuals and public institutions alike. Many homes only have access to their data at the mercy of their supplier. Often data is presented as standard graphs and the raw data cannot be accessed or shared easily.
Third parties have to undergo extensive legal and administrative procedures to gain access and verify that they have been given consent by a household to access their data. As a result, few provisions are available for households to access their own data and these services tend to come at a cost, which is prohibitive for many vulnerable homes.
The roll-out of smart meters is motivated by clearly identified benefits to utilities. Remote meter readings, reduced mis-billings, and the ability to remotely disconnect customers (or move them to a pre-payment meter) result in cost savings for utilities. The benefits for end users are less well understood and the claimed energy savings are greeted by many end-users with suspicion. A better case for the public benefit of smart meters needs to be made to build trust and willingness to share data for public and private good.
Aims of the research
This studentship will systematically research the privacy implications of smart meter data from a social and a technical perspective. Smart meter data is rightly considered sensitive. Even anonymised data can contain potentially discriminating features, such as religion (e.g. changes in use patterns during Ramadan or other festivities). Heating patterns can reveal underserved heating needs, which could result in respiratory ill health. Such data can be used to target retrofit measures, but it could equally be used by health insurance companies to disadvantage already fuel poor households.
Nuanced considerations of a social and a technical nature need to be weighed up to inform policies on the handling and access to energy related data.
This doctoral thesis will address the following research question:
How can smart meter data be processed and shared for the benefit of people and climate goals?
This overarching question is broken down into three sub-questions:
What are the specific privacy and discrimination concerns for different user groups?
What data features are needed by different stakeholder groups?
What measures and processes are effective at safeguarding data privacy while maintaining valuable features?
Express your interest in the opportunity
Background
Much has been written about the need for ‘open data’ (Energy Digitalisation Taskforce, Open Data Institute and others). While the benefits of better access to data are widely agreed upon, the nature of the potential downsides is often discussed without direct involvement of citizens themselves. As originators and ‘owners’ of much of the end-use related energy data it falls to them to give ‘informed’ consent. The ‘data controllers’ (in GDPR speak) barely know what power smart meter data holds and to what extent it may reveal unexpected personal and sensitive insights about people. It is therefore a big ask to expect citizens, many of them potentially vulnerable, to give ‘informed’ consent about the use of their smart meter data. How much exposure are they being asked to consent to?
The current smart meter procedures stipulate that smart meter data does not pose a disclosure risk so long as a sample of at least 10 meters is aggregated. The number ten is cautious, but arbitrary. There is no scientific literature or detailed explanation of the limits or the nature of the disclosure risk.
Load data is highly diverse. Extreme load patterns may still stand out, even within an aggregated sample. At the same time, much of the temporal detail and diversity of use patterns is lost in aggregates, reducing their value for researchers and analysts.
This studentship we will test the limits of privacy preservation, through aggregation and differential privacy, analytically and experimentally. We have developed novel load clustering approaches using Dynamic Time Warping, which form profile groups with similar load patterns and common explanatory variables, such as appliance ownership and a rich set of socio-demographic survey information and activity diaries. Aggregating by archetypes can preserve some of the relevant ‘signatures’ of profiles and improve their usefulness. We don’t yet know what the privacy implications of these methods are.
Building on these signatures as benchmarks for ‘realistic’ patterns, this studentship will apply new methods to generate synthetic data. These synthetic profiles will transform the extent to which household-level data can be made available for research and practical applications. They contain all relevant characteristics from their training data, but cannot be directly attributed to any individual.
Archetype and synthetic profiles can be made publicly available, with no need for costly, labour and time intensive security procedures. Anyone can access such data for analysis or to compare groups - including ones own profile with ‘people like yourself’. This last feature opens up unique opportunities for personalised feedback and advice as a new data-driven service.
Novelty
The novelty of this thesis is in its unprecedented access to smart meter data and explanatory meta data, made possible by the EDOL project.
This opportunity will lead to three major advances over previous theoretical deliberations about privacy:
This thesis will establish a scientific basis that can guide policy and regulation on the handling of energy relevant data, replacing arbitrary privacy rules with evidence led conditions for disclosure.
The contextual socio-demographic information allows to establish the potential for discrimination and identification at a level that has hitherto not been possible with smart meter data alone.
This thesis takes a citizen-centric approach and focusses on the benefits and costs for end-users, rather than merely the system or utilities
Research methods
This project is made possible as part of the Energy Demand Observatory and Laboratory (EDOL EP/X00967X/1), for which Oxford leads on data analysis. The doctoral student with be co-supervised by Dr Phil Grunewald (EDOL’s technical lead) and Prof. Tina Fawcett (EDOL’s social lead), and embraces the inter-disciplinary breadth for their work. EDOL is funded until the end of 2027, such that part-time and flexible arrangements can be supported.
EDOL will provide the wealth of data required for this undertaking. We also seek to facilitate access to key learnings on social and technical considerations, which the supervisors teach on Oxford’s Energy Systems MSc.
We have established methods for handling and processing personal and sensitive energy time-series and survey data. The doctoral student will be able to use these and adapt them as part of their studies.
As part of a major research programme, the doctoral student is expected to benefit from a stimulating environment of data-science, social-science and energy experts, who are working on closely related and mutually relevant topics (see University support below).
The methods will be more problem-focussed than theory-based:
Extensive engagement with the UK energy data community (DCC,Energy Digitalisation Taskforce…) research programmes (CREDS II, SERL, UKERC, EnergyREV…) and literature on energy data, data privacy and ‘smart’ systems.
Data analysis: use available smart meter data to test aggregation, differential privacy and clustering approaches for their effectiveness to sustain or obfuscate data content.
Citizen engagement: Surveys and/or interviews with representative participant groups on privacy perceptions, concerns and appreciation of the extent of inferences that can be made from energy data, using examples from the data analysis and aggregation approaches above.
By the end of their thesis, the student will have acquired energy data skills of great value to the UK and it’s zero carbon ambitions. They furthermore will have developed meaningful connections across a wide range of energy research disciplines and programmes.
This doctoral thesis will strengthen CREDS by making multivariate socio-technical data available to more researchers. It is expected that appropriate privacy protocols will make more data with more analytical value openly accessible with fewer restrictions and procedures.
Of particular relevance to CREDS is the opportunity to link large sources of data between research themes, such as transport and flexibility, combine them with supply side programmes, and produce a strong evidence base for policy advice.
University support and research environment
Oxford has a vibrant and thriving energy research community. A fast growing number of postgraduate and doctoral students in energy studies adds to this community and creates a lively and stimulating environment for exchange and debate. Energy research is organised formally and informally with the support of numerous networks, collaborations and dedicated institutions.
The overarching networks are the Oxford Networks for the Environment (ONE), consisting of Food, Climate, Biodiversity, Water and Energy Networks.
Dr Grunewald was instrumental in the creation of the Oxford Energy Network. Under the leadership of Prof Sir Chris Llewellyn Smith and Prof Nick Eyre this network grew to over 200 senior researchers from more than 20 Oxford departments. Dr Grunewald later led the network himself and remains involved as a member of the steering committee. The network is an inter-disciplinary hub for researchers from across the University and provides regular seminars, which are well attended by senior academics and doctoral students alike. The annual energy conference has been held by the network for over a decade now.
Oxford is also home to many student led initiatives, including the Oxford Energy Society, of which Dr Grunewald is the Senior Member. The committee of students host weekly talks with distinguished external speakers, followed by pizza and networking.
Support at the Department of Engineering Science
At department level, students link into existing equality and diversity support structures: each department has a nominated Disability Contact, and a Harassment Advisor of each sex; Oxford is a Stonewall Diversity Champion; and the University holds Athena SWAN Bronze status, with three Silver and seven Bronze department awards. Policy and practice are co-ordinated by a central Equality & Diversity Unit. 3 Students will receive further support and academic benefits – at no cost to EPSRC – through the Oxford college system. Membership of a vibrant academic community drawn from across the sciences, social sciences and humanities exposes students, through informal peer-to-peer contact, to a far broader range of ideas and ways of thinking.
College support
Oxford colleges provide a caring and supportive environment, which is especially helpful for doctoral students, who otherwise lack the network that naturally comes with taught coursed. Successful applicants will be encouraged to join Oriel College, where Dr Grunewald leads the Energy Theme of the Oriel Environmental Group and where students can connect with other students who currently undertake doctoral studies in energy related topics. Prof. Nick Eyre and Prof. Charlie Wilson are Fellows at Oriel College, giving the college a world-class standing in energy research and attracting students with a keen interest in energy and the environment.
Research groups
Oxford is home to several interdisciplinary research groups with particular focus on social and technical issues in energy. Dr Grunewald is connected to many of them through collaborative work, which this studentship stands to benefit from.
The long established Energy Group in the Environmental Change Institute (ECI) is well known for its work of fuel poverty (Brenda Boardman), energy efficiency (Nick Eyre), citizen engagement (Sarah Darby) and energy policy (Tina Fawcett). CREDS has attracted many highly talented researchers to the ECI and the continue to have a major impact on EDI and equity in energy research.
Dr Grunewald developed his EPSRC Fellowship on an activity based understanding of energy demand in the Energy Group and maintains collaborations with several group members, including the group leaders Prof. Tina Fawcett (joint Oxford Martin School funding) and Prof. Charlie Wilson (joint EPSRC grant).
Dr Grunewald is directly involved with two research groups in Engineering Science. The Battery Intelligence Lab under leadership of Prof. David Howey develops advanced machine learning approaches for the understanding of battery health. As part of our cross-disciplinary engagement, it was possible to apply the Dynamic Time Warping methods, developed by this group, to the classification of household energy demand profiles. This work forms part of the studentship research and leading researchers in the Howey group can provide support and expertise.
Dr Grunewald is also part of the Energy and Power group, led by Prof. Malcolm McCulloch. This groups addresses a wide range of energy access and local energy issues with a highly interdisciplinary team.
Doctoral students therefore enter a vibrant, stimulating and supportive community of energy demand researchers.
MSc in Energy Systems
Dr Grunewald developed the Oxford proposal for an MSc in Energy Systems, which builds on Oxford’s strengths in inter-disciplinary whole-system research. This course attracts over 200 applications each year and delivers content spanning social, technical and strategic challenges in energy. The material taught on this course is extremely valuable to doctoral students in the energy space and students are encouraged to attend selected lectures.
ZERO institute
Oxford has recently made a strategic £3.25m investment in energy research with the establishment of the Zero-carbon Energy Research Oxford (ZERO) institute, which has zero carbon energy use as one of its three main strategic themes. A Chair has been created to lead this institute and six academic positions, including two in energy use, will be advertised shortly, further strengthening Oxford’s position as a leader in energy demand research.
The ZERO institute provides an excellent environment for doctoral students to engage directly with fellow researchers across a wide range of disciplines.
Doctoral Training
Specialised 1-day transferable skills courses are offered by the Graduate School of the Oxford’s Mathematical and Physical Life Sciences (MPLS) Division. Courses are typically available twice per year. These include poster design, thesis and report writing, conference (choosing and networking), interview skills, and coaching on building a business. The full postgraduate training and development programme is available on the Information DPhil Students mpls.ox.ac.uk/training/pgr/PGR website.
