Edge Artificial Intelligence for Domestic Energy Efficiency

With adverse industrial effects on the global landscape, climate change is imploring the global economy to adopt sustainable solutions. Domestic households, for example, contribute to almost a third of overall energy consumption in the European Union. Hence, the ongoing evolution of energy efficiency targets massive data collection and Artificial Intelligence (AI) for big data analytics. Besides, the emergence of edge computing is playing a rising role in deploying AI solutions on resource-constrained platforms, while recommender systems continually impact multiple verticals by introducing automated intelligence to decision-making. This presents an opportunity of addressing incumbent energy efficiency challenges using recent advances in edge computing and AI. Therefore, this thesis presents a framework for improving energy efficiency in domestic buildings through data collection, classification, and recommendations with emphasis on edge computing capabilities. The contributions of the work can be summarised as follows: First, a novel dataset is produced from a UK-based household that comprises appliance-level energy and ambient environmental data. Second, a novel method for visualising and summarising time series energy data is devised using Gramian Angular Fields (GAF), to produce pictorial representations for further classification on High-Performance Edge Computing (HPEC) platforms. Third, a modular recommender system is developed to produce personalised energy-saving recommendations on HPEC platforms. Machine Learning (ML) has been used for data classification and the recommender system. Specifically, a lightweight transfer learning model is employed to detect anomalies in GAF data and is tested on four HPEC platforms achieving near-real-time computational performance. The system is collectively evaluated on the HPEC platforms based on standard metrics, computational time, and potential impact on overall energy savings. Results demonstrate performance of up to 17.5 msec/GAF image for classification, 29.13 sec for generating recommendations/month, and projected cost savings of 10% of annual electric bills for a typical UK household.

Name dc:type.qualificationname

PhD

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

De Montfort University

Year dc:date.issued

2023

Author dc:creator

• Alsalemi, Abdullah

Licence

• https://dora.dmu.ac.uk/bitstreams/8e5e877a-9a7b-46e8-a803-0032079c6404/download