Stochastic modelling and genetic algorithm-based optimal control of air conditioning

There has been widespread concern over the high energy consumption and the often less-than-satisfactory environmental control performance of most air conditioning systems relying on conventional control schemes. In this paper, a new approach to tackle the problem is presented, which aims at achieving a high quality control of the indoor thermal environment with reduced energy consumption. The rationale of the approach is to employ an optimization procedure for the control of an air conditioning system to achieve specified thermal comfort levels with minimum power input based on forecasts obtained from a stochastic mathematical model of the system. Such a model, which embodies the interaction amongst the internal and external environments, the building structure and the air conditioning system, can be derived using time-series analysis of the actual performance data and the environmental conditions of the exterior and the conditioned space. The power consumption of the air conditioning system, indoor and outdoor air temperatures, and solar radiation are the variables included in the present time-series model. The model is utilized to develop an optimal control strategy based on genetic algorithms. In this study, the power consumption of the air conditioner is chosen to be the objective function to be minimized using genetic algorithms against specified comfort levels based on Fanger's thermal comfort index of predicted mean vote (PMV). Computer simulation is used to compare the performance of the optimal controller using genetic algorithms and that of on-off and PI controllers. The results show that a reduction in power consumption has been achieved in most cases by the optimal controller while maintaining a higher degree of thermal comfort.