Occupant-Centric Energy Management for Small Commercial Buildings

Ande, Ruth (2020) Occupant-Centric Energy Management for Small Commercial Buildings. Doctoral thesis (PhD), Manchester Metropolitan University.

Preview

Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (3MB) | Preview

Abstract

As the UK strives to reduce its impact on the environment, small and medium sized enterprises (SMEs) face significant energy reduction barriers which include high costs, the lack of expertise and significant time limitations. Many energy management systems (EMS) do exist but they are largely inaccessible to SMEs because they generally fit into three categories: being complex and expensive; affordable but requiring expertise to fit and manage; or affordable but overly simple and ineffective. Therefore, this thesis focuses on the development of on an holistic occupant-centric EMS to overcome the limitations of existing solutions to enable SMEs to overcome the barriers they have experienced. The principle of the occupant-centric EMS is to improve the temporal match between building occupants and energy consuming systems. To meet this principle, a number of enabling technologies are utilised including, Internet of Things (IoT), wireless sensor networks (WSN) and machine learning (ML). The major contributions of this work include the development of • a WSN simulation tool • a methodology to analyse different network deployment techniques • creation of a large labelled multimodal data set • a single mode and multimode ML architecture which is designed and deployed on a constrained edge-based system to utilise binary classification to determine occupancy • a holistic low cost occupant-centric EMS which automates a significant reduction of energy consumption within small commercial buildings A number of node placement algorithms are developed to assess existing WSN deployment techniques that are utilised for unobtrusive, privacy protecting IoT data capture. The most suitable technique is determined to be the sensor grid which uses 44% of the hardware of other deployments and demonstrates an accuracy of 81% for occupancy monitoring. To further improve the performance of occupancy monitoring, an edge-based ML model which analyses thermal image data is designed and iii implemented demonstrating more than 96% accuracy in an office environment. To improve the performance in a wider range of environments, the ML model is extended to enable simultaneous analysis of the IoT multimodal building data. This model achieves the same performance in the office but demonstrates a 15% improvement in sensitivity and 31% in precision in another environment. The utilisation of additional low cost sensors and data fusion techniques enable an increase in building coverage from 78% to 100%, whilst maintaining the quantity of IoT nodes. The completed developed occupant-centric IoT-based EMS costs less than a fifth of existing comparable systems. The experimental evaluation results demonstrate more than 10% reduction in total building energy consumption whilst maintaining a comfortable working environment.