Abstract
dc:description.abstractThe transition to an intelligent, reliable and efficient smart grid with a high penetration of renewable energy drives the need to maximise the utilisation of customers’ demand response (DR) potential. More so, the increasing popularity of smart meters deployed at customers’ sites provides a vital resource where data driven strategies can be adopted in enhancing the performance of DR programs. This thesis focuses on the development of new methods for enhancing DR in smart grids using big data analtyics techniques on customers smart meter data. One of the main challenges to the effective and efficient roll out of DR programs particularly for peak load reduction is identifying customers with DR potential. This question is answered in this thesis through the proposal of a shape based clustering algorithm along with novel features to target customers. In addition to targeting customers for DR programs, estimating customer demand baseline is one of the key challenges to DR especially for incentive-based DR. Customer baseline estimation is important in that it ensures a fair knowledge of a customers DR contribution and hence enable a fair allocation of benefits between the utility and customers. A Long Short-Term Memory Recurrent Neural Network machine learning technique is proposed for baseline estimation with results showing improved accuracy compared to traditional estimation methods. Given the effect of demand rebound during a DR event day, a novel method is further proposed for baseline estimation that takes into consideration the demand rebound effect. Results show in addition to customers baseline accurately estimated, the functionality of estimating the amount of demand clipped compared to shifted demand is added.
Degree
thesis:*- Level dc:type.qualificationlevel
- PhD
- Grantor dc:publisher.institution
- University of Glasgow
- Year dc:date.issued
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Oyedokun, James Timilehin
Subjects
dc:subject × 1Rights
- Language dc:language
- en
Identifiers
dc:identifier.*- DOI dc:identifier.doi
- https://doi.org/10.5525/gla.thesis.82616