Best practice for modelling uncertainty in usage patterns?
I am trying to capture uncertainty due to user behaviour with my model and wondered if anyone had any suggestions for best practice on how to do this?
I am carrying out multiple runs to explore the effects of uncertainty around usage patterns (occupancy, lighting, equipment and heating cooling set point schedules). This is part of a global sensitivity analysis and the more variables I have the more runs I will need.
The two approaches I have considered are:
- Define a range of different schedules for each category and randomly select a schedule in each category for each model run. This would allow me to have relatively complex schedules but the consideration of uncertainty is limited to the range captured in the schedules I have defined. This would only require one variable per schedule type.
- Include a range of variables in each schedule to simulate changes in state. If the schedule is simple (eg. occupancy is zero until a start time, then 1 until an end time) then this is relatively straightforward and only needs 2 variables but the number of variables increases if the schedules is more complicated (eg. off for an hour at lunch-time).
At the moment, I think my preferred approach is the 2nd one, with a simplified schedule, so just a variable for the on time and a variable for the off time (ie. 2 variables per schedule). However, I haven't been able to find any precedents in the literature to suggest an approach. Does anyone have any suggestions for a better approach?
Thank you!
