A range of different scenarios have been predicted for future UK energy supply. While there is significant uncertainty, all expect an increase in small-scale distributed generation integrated in constrained or independent networks and with predominantly domestic consumers. This reduction in system scale has not, however, driven a significant change in design practices, with deterministic models and rules-of-thumb prevalent. Little consideration has been given to how the specific household characteristics and the size of system impact on demand level and timing, the degree of uncertainty in anydemand prediction, and how design practices should change to reflect this. The main contribution of the presented work has been to address this. To allow the variation and uncertainty to be quantified; a highly differentiated, probabilistic, bottom-up demand model has been developed for electrical and hot water use. The 1-minute resolution model incorporates an enhanced Markov chain occupancy model and is based on a newly developed discrete-event approach for occupant-initiated demands. Utilising realistic factoring for appliance ownership, income, occupancy, and random energy-use behaviours, the model has been shown to capture the range of potential household demands. Assessment that the developed model, and any existing model calibrated using group data, tended to rapidly converge to the group average basis, prompted further method development to improve the model's performance in capturing individual household demand behaviours. Analysis of both existing data and the demand model output has shown that energy system demand can vary significantly based on socio-economic characteristics and the types of households supplied. It also highlights that demand uncertainty for individual households can exceed an order of magnitude, even if household characteristics are known. As the system scale is increased, the level of overall demand uncertainty remains significant to at least 200 household systems. A method has therefore been developed that allows multiple runs of the probabilistic model to be reduced to a representative subset, which can be used to analyse potential energy system performance scenarios probabilistically using existing optimisation tools.
|Date of Award||25 Apr 2017|
- University Of Strathclyde
|Sponsors||BRE Trust & University of Strathclyde|
|Supervisor||Nicolas Kelly (Supervisor) & Joseph Andrew Clarke (Supervisor)|