Stochastic modelling of building heating and cooling systems

Abstract

Computer simulations, which can model the operation of the heating, ventilation and air conditioning plant in buildings, generally have lengthy execution times when they are used to generate long-term measures of performance. This investigation has established the feasibility of using a stochastic model, based on Markov chains, wich preserves the dynamic nature of the problem but offers reduced execution time. There is a trade-off between reduction in execution time and the accuracy of the simulation. A simulation enabled a simple combined building/plant model to be run in both Markov and conventional (time-series) modes, thus permitting a direct performance comparison. It was found possible to obtain significant reduction in execution times by running a Markov model without incurring unacceptable inaccuracies. The technique was considered to be worthy of further development in the context of plant simulation.