PhD in Distributed Machine Learning for Privacy-Preserving Non-Intrusive Load Monitoring

About the Project

Project details:

Smart meters have become increasingly common in UK homes, yet most only report total consumption, offering little insight into which appliances drive demand. Non-Intrusive Load Monitoring (NILM) can bridge this gap by turning whole-home readings into appliance-level information, with studies showing meaningful efficiency gains for households. However, deploying NILM at scale raises privacy and real hardware constraint concerns. 

This PhD will focus on those challenges by developing a distributed, privacy-preserving NILM framework, so we can move from small research pilots to real-world systems. The overarching aim is to deliver a scalable approach, pairing shared “aggregator” models with household-specific “client” models that exchange knowledge while keeping data private.

This PhD will focus on three strands of work: 1) Innovate NILM model structures. Design efficient neural network architectures for both aggregator and client models that meet strict accuracy-efficiency trade-offs, using automated configuration to find Pareto-optimal designs under real deployment constraints. 2) Build the distributed learning loop. Develop the learning and update mechanisms that coordinate many clients via regional aggregators, including federated learning steps that let homes benefit from each other’s patterns without exposing local data. 3) Prototype and validate in practice. Deploy the framework in an industrial testbed and evaluate accuracy and computational performance under realistic load conditions. 

The student will work closely with Chameleon Technology, the UK market leader in energy display and smart-energy services, who bring deep domain expertise and a truly practical environment. The studentship includes access to a comprehensive UK-wide whole-home energy dataset with rich tagging for model training, a fully equipped testbed with real smartmetering devices installed in homes, and dedicated engineering time to mirror production settings. The student will collaborate with engineers through regular technical reviews and integration sprints, gaining first-hand experience of how research transitions into deployable services.

By the end of the project, the student will deliver optimised NILM architectures, a distributed learning framework, and an industrially validated prototype, backed by publications, codes, and a clear path to further development and commercialisation opportunities. The project provides a chance to shape deployable AI for smarter, more efficient energy use, working from first principles all the way to real-world impact. 

Please direct project specific enquiries to: Zhou Zhou (z.zhou2@exeter.ac.uk)

Please ensure you read the entry requirements for the potential programme you are applying for. To Apply for this project please click the 'Apply' button  

Funding Comment 

Payment of tuition fees (Home), Research Training Support Grant £5,000 over 3.5 years 

Location: Harrison, Streatham

Advert information