Indoor Air Conditioning Based On Model Predictive Control

With the development of modern intelligence house, people pay more and more attention to high-quality indoor air-conditioning systems. Basic functions of cooling and heating could not satisfy people as usual; Indoor thermal comfort is advanced. So the design and control of air-conditioning systems need more research.Characteristics of air-conditioning system are dynamic, time-varied, nonlinear, large time-delay, random disturbance, serious coupled. Traditional PI/PID control and modern control aiming at MIMO linear time-invariant system, even some intelligent control is insufficient. The main difficulties are: hard to build exact mathematical models; unable to completely eliminate the influence of uncertain disturbances; decoupling is uneasy and the control effects will be worse because of unsuccessful decoupling. By analyzing the influent factors of indoor air environment, this paper constructed the mathematical parameter models of air-conditioning system and identified them via least square sequential algorithm, then brought out a feedforward-decoupling generalized predictive control algorithm. This algorithm formed a closed-loop structure by combining feedforward-decoupling compensator with GPC controller and effectively eliminated the coupling among loops. Besides, for GPC methods just need simple models, this algorithm used multi-step prediction and rolling optimization, enriched the information of future changing trend, so as to overcome various influence of uncertain disturbance and complex variation.The regulating effects of traditional air-conditioning are to cool, heat, ventilate and keep moisture. Further research of thermal comfort and indoor air quality bring in comfort control, a new concept to the air-conditioning control field. This paper brought this concept into the control strategy referred above, comprehensively analyzed the chief factors influencing indoor thermal comfort environment, real-time calculated thermal comfort index (PMV), then constructed PMV index self-tuning outer loop via least square algorithm, so as to form a double-closed-loop control algorithm based on feedforward-decoupling generalized predictive control and self-tuning of PMV index. The algorithm took human thermal comfort index as ultimate control goal, so it can obtain conditioning results as"human-oriented". Simulation results indicate the good performance of this control algorithm.