Integrated modelling of electrical energy systems for the study of residential demand response strategies

Sancho Tomás, Ana (2017) Integrated modelling of electrical energy systems for the study of residential demand response strategies. PhD thesis, University of Nottingham.

[img]
Preview
PDF (Thesis - as examined) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (19MB) | Preview

Abstract

Building and urban energy simulation software aim to model the energy flows in buildings and urban communities in which most of them are located, providing tools that assist in the decision-making process to improve their initial and ongoing energy performance. To maintain their utility, they must continually develop in tandem with emerging technologies in the energy field. Demand Response (DR) strategies represent one such family of technology that has been identified as a key and affordable solution in the global transition towards clean energy generation and use, in particular at the residential scale.

This thesis contributes towards the development and application of a comprehensive building and urban energy simulation capability that parsimoniously represents occupants' energy using behaviours and responses to strategies to influence them. This platform intends to better unify the modelling of Demand Response strategies, by integrating the modelling of different energy systems through Multi Agent Simulation, considering stochastic processes taking place in electricity demand and supply. This is addressed by: (a) improving the fidelity of predictions of household electricity demand, using stochastic models, (b) demonstrating the potential of Demand Response strategies using Multi-Agent Simulation and machine learning techniques, (c) integrating a suitable model for the low voltage network to study and incorporate effects on the grid, (d) identifying how this platform should be extended to better represent human-to-device interactions; to test strategies designed to influence the scope and timing of occupants' energy using services.

Stochastic demand models provide the means to realistically simulate power demands, which are subject to naturally random human behaviour. In this work, the power demand arising from small household appliances is identified as a stochastic variable, for which different candidate modelling methods are explored. Variants of two types of stochastic models have been tested, based on discrete time and continuous time stochastic processes. The alternative candidate models are compared and validated using Household Electricity Survey data, which is also used to test strategies, informed by advanced cluster analysis techniques, to simplify the form of these models.

The recommended small appliance model is integrated with a Multi Agent Simulation (MAS) platform, which is in turn extended and deployed to test DR strategies, such as load shifting and electric storage operation. In the search for optimal load-shifting strategies, machine learning algorithms, Q-learning in particular, are utilised. The application of this new developed tool, No-MASS/DR, is demonstrated through the study of strategies to maximise the locally generated renewable energy of a single household and a small community of buildings connected to a Low Voltage network.

Finally, an explicit model of the Low Voltage (LV) network has been developed and coupled with the DR framework. The model solves for power-flow analysis of a general low-voltage distribution network, using an electrical circuit-based approach, implemented as a novel recursive algorithm, that can efficiently calculate the voltages at different nodes of a complex branched network.

The work accomplished in this thesis contributes to the understanding of residential electricity management, by developing better unified modelling of Demand Response strategies, that require integrated modelling of energy systems, with a particular focus on the study of maximising locally generated renewable energy.

Item Type: Thesis (University of Nottingham only) (PhD)
Supervisors: Robinson, Darren
Sumner, Mark
Keywords: demand response; renewable energy; modelling; machine learning; community energy; low voltage network
Subjects: N Fine Arts > NA Architecture
T Technology > TJ Mechanical engineering and machinery
Faculties/Schools: UK Campuses > Faculty of Engineering > Built Environment
Item ID: 46872
Depositing User: Sancho Tomás, Ana
Date Deposited: 13 Dec 2017 04:40
Last Modified: 02 May 2018 03:29
URI: https://eprints.nottingham.ac.uk/id/eprint/46872

Actions (Archive Staff Only)

Edit View Edit View