The project treats heating as a planning problem: predict enough of the building’s near future to choose a useful control action, then solve again as new information arrives. A two-state RC model, constrained SLSQP optimization, stochastic weather inputs, and repeated analysis keep the comparison between thermostat and MPC grounded in measurable behavior.
PROJECT NOTE2026 · Simulation · Research project
Predictive Heating MPC
A stochastic building simulation comparing a thermostat with constrained model-predictive control.