Modelling of individual domestic occupancy and energy demand behaviours using existing datasets and probabilistic modelling methods

With growing trends towards smaller-scale, low carbon energy systems such as microgrids and later generation district heating, coupled with the increased use of building simulation in bottom-up stock modelling, there is increasing interest being able to automatically generate multiple, high-resolution profiles both occupancy occupancy-dependent household demand. This has resulted emergence a range tools capable producing time-varying demand that capture many characteristics evident real-world data. However, result limited data availability, these have typically been calibrated using composite from multiple individuals or households. results production which whilst statistically faithful average underpinning data, fail variance seen household-to-household real world. paper attempts address this shortcoming, developing testing new method within an existing model structure for linked probabilistic representative overall population it derived, but improved ability specific outputs better match behaviour The approach utilises previously established modelling methods, along population-based calibration used models. temporal predictive basis models manipulated account individual behaviours retaining statistical source was achieved by adapting Markov chain timing developed factoring probability values. A electrical also adapted manipulation relative different appliances demands differences between behaviours. described benefit not requiring any additional often scarce. predictions previous version are compared indicating alterations enable significantly more diverse be generated, still supporting