الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 120 |  |  |
| | | from | 120 | | |
Abstract
The work presented in this thesis addresses the problem of the rapid energy demand in the residential sector. With the advent of the IoT, governments have been paying much attention to build up the smart grid environment to meet the needed demand.
This work presents a methodology for extracting interesting patterns from data generated by smart meters. Extracting these patterns is a challenging task. However, it has a great benefit for both utilities and home residents. For utilities, several applications can be developed for power distribution. For home residents, the more they understand their behavior the better control they have over their consumption usage.
The proposed work extracts home residents preferences through discovering appliance-time association and appliance-appliance association. The appliance-time association reveals the appliances that are preferred to be used at a time. The appliance-appliance association reveals the appliances that are preferred to be used together at a time.
The proposed work mines smart meter data incrementally at the end of each day without mining the whole dataset. The basic idea of the proposed work is to extract residents behavior taking into account the usage time and priority. The UTARM algorithm has been extended in this work for supporting these two factors: temporal factor which was the hour and utility factor which was the weight for using an appliance at the hour.
The UTARM algorithm has been used as an initial phase in the proposed methodology. The findings of this phase has been used to extract appliance-time association in addition to appliance-appliance association by generating association rules and hierarchical clustering.
The main difference between hierarchical clustering and association rules is that the association rules extract the exhibition period for each discovered association per each hour thereby a history of residents behavior can be obtained. The hierarchical clustering finds associations across the 24-hours for the whole recorded days.
The UK-DALE dataset has been used to evaluate the proposed work. It holds consumption logs for five dwellings with different durations for 4.3 years.