A disaggregated, probabilistic, high resolution method for assessment of domestic occupancy and electrical demand

Abstract

An integrated domestic occupancy and demand model with a 1-min resolution has been developed which better captures the influence of different occupant behaviours than previous models. The occupancy model includes the fundamental link between occupancy and demand, and differentiates between different types and sizes of households. In particular, the likelihood of daytime occupancy is captured by age and employment differentiators. A novel method for identifying appliance use events and linking use to an occupancy profile has been developed that accounts for household specific appliance usage using an event-based approach calibrated directly from measured data. The method has been shown to perform better than both per-timestep probability models and models calibrated from time-use survey activity diaries. To further capture individual household behaviours, the demand model incorporates additional factoring to account for income and random behavioural influences. Whilst improving differentiation of individual household energy usage, due to limitations in the available data, the model incorporates some occupancy and use behaviour factors that are a composite of multiple households, leading to some behaviour averaging in the model output; consequently the model is best employed for energy demand assessment of multiple households

A cross analysis of existing methods for modelling household appliance use

“…In contrast to that, TUD based models are capable of producing inter-household variation generated by socio-demographic conditions. However, Flett and Kelly (2017) revealed that the inter-household variation that can be generated considering socio-demographic conditions is smaller than the variation observed in empirical data.…”

“…To replicate the intra/inter-household variation in TUD based models, factors determining intra/inter-household variation should be taken into account. TUD based models could be improved by using longitudinal time use data or taking into account occupants' weekly schedules such as the distribution of the number of uses of an appliance throughout the week (Flett and Kelly, 2017).…”

“…One of the weaknesses of Richardson's model is that the variation in the number of switch-on events per day cannot be replicated because switch-on events occur as a result of random trials made at each time step. Flett and Kelly (2017) overcome this weakness by first determining the number of switch-on events on the simulated day based on empirical data. The switchon events are then allocated to the timeline by considering occupancy.…”

“…This result implies that the TUD based models cannot reproduce intra-household variation. This is because the TUD based models only reproduce population means Flett and Kelly (2017). demonstrated that TUD based models can be improved in the replication of the -on events per day by adopting the above-mentioned approach ofYilmaz et al (2017).…”

Adaptation and validation of an existing bottom-up model for simulating temporal and inter-dwelling variations of residential appliance and lighting demands

“…Bottom-up AL models may be further categorized as either explicit or implicit-occupancy models. Explicit-occupancy models, referred to by Flett and Kelly (2017) as occupancy-to-demand models, are driven by inputs or estimates of explicit occupant presence in dwellings at each timestep. Alternatively, implicit-occupancy models often rely on AL usage statistics to determine when AL devices are turned on.…”