Real-time and semantic energy management across buildings in a district configuration

Reynolds, Jonathan ORCID: https://orcid.org/0000-0001-9106-9246 2019. Real-time and semantic energy management across buildings in a district configuration. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Download (8MB) | Preview
	PDF - Supplemental Material Restricted to Repository staff only Download (168kB)

Abstract

Existing building and district energy management strategies are in urgent need of an overhaul to meet the energy and environmental challenges of the 21st Century. The immense growth in the availability of data through the Internet of Things (IoT), the decentralisation of energy generation, and the increasing power of Artificial Intelligence (AI) presents an opportunity to achieve a paradigm shift in the way energy is controlled and managed. To contribute to this field, this PhD project undertook a thorough literature review combined with a participatory, action research approach to identify and understand the key challenges faced by facility managers and to identify potential areas of improvement. Following this, the PhD thesis aims to tackle three key research areas using simulated case study experiments. These aim to optimise thermal energy management within buildings at a zone-level, control energy generation at a district-level, and combine the learnings from these two experiments with a holistic energy management solution that controls both the energy supply and demand at a building and district-level. At a building-level, a model predictive control approach combining a genetic algorithm and surrogate artificial neural network is used. A predictive and context aware controller is able to produce 24 hour heating set point schedules for each zone within a building. This approach achieved an energy saving of 18% whilst maintaining thermal comfort for users. The methodology also had the capability to adapt to dynamic energy pricing tariffs and capable of optimising for energy cost by shifting load to cheaper periods. At a district-level, a predictive, optimisation-based approach was developed to determine the operation of a multi-vector, district heating, energy centre. When thermal storage and several generation sources are available, alongside variable renewable energy generation and building demand, static, rulebased controllers cannot perform adequately in all conditions. Instead, the optimisation-based approach, developed in this thesis, was able to increase profit to the energy centre by 45% as well as decrease CO2 emissions whist adapting to errors in energy demand and supply forecasting. Finally, the most significant contribution of this thesis was provided by efvii fectively combining the approaches made at a building and district-level. This case study aimed to simultaneously control the energy generation of the district energy centre, alongside the thermal demand of one of the buildings within the district. The additional flexibility provided by partially controlling the building demand led to a further 8% increase in profit to the energy centre, compared to just optimising energy supply. This demonstrates the vital importance of treating the consumer as an integral, active component of the energy system. It is argued that the contributions made throughout this thesis will become more relevant when coupled with additional research fields. This includes the growth in available data from IoT sources, advanced AI including unsupervised learning, and utilising a shared semantic description of smart building, smart energy and smart city concepts. At its core, this thesis aims to demonstrate that ‘thinking’, predictive, control strategies, that are more context-aware, can achieve significant benefits over the traditional reactive, rule-based controllers of the past.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Engineering
Uncontrolled Keywords:	Multi-Vector Energy Systems; Building Energy Management; Renewable Energy; Genetic Algorithms; Machine Learning; Model Predictive Control.
Date of First Compliant Deposit:	2 April 2019
Last Modified:	04 Nov 2022 11:54
URI:	https://orca.cardiff.ac.uk/id/eprint/121237

Actions (repository staff only)

Edit Item