Demand Response Fleet Model: Thermostat Population Bin Dynamics (2D)
Interactive 2D population-density (bin transition) model of a demand-response fleet of thermostats: watch the temperature distribution of on/off houses circulate around the dead-band hysteresis loop, and see how direct load control -- and whether the restart is staggered or synchronized -- shapes the aggregate grid load and the post-event cold-load-pickup rebound.
This 2D companion to the 3D demand-response simulator replaces the per-house agent grid with the population-density ("state-bin") model actually used in demand-response research: two densities over temperature — one for houses currently off, one for houses currently on — advect along the same dead-band hysteresis ODE as the 3D sim's individual houses, circulating mass around a closed loop instead of tracking each house separately. Direct load control opens and shuts a gate at the switching boundary for enrolled cohorts, diverting their outflow into a backlog reservoir that keeps warming until its cohort's own schedule releases it. Watch the histogram show the population's temperature distribution shift live, and compare a rotating (staggered) restart against a synchronized one to see, in the aggregate load curve, why staggering is the real-world defence against a cold-load-pickup rebound.
A 2D population-density (state-bin) model of the same demand-response thermostat fleet as the 3D twin: two densities over temperature -- one for houses off, one for houses on -- advect along the real dead-band hysteresis ODE and circulate mass around the hysteresis loop as a finite-volume upwind scheme, instead of tracking each house as a separate agent. Direct load control opens and shuts a gate at the switching boundary per phase-offset cohort, diverting curtailed outflow into a backlog reservoir; watch the population histogram shift live and compare a rotating (staggered) restart against a synchronized one to see why staggering tames the cold-load-pickup rebound.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install