Multi-objective control of central air conditioning system

A practical and data-driven multi-objective particle swarm optimization (MOPSO) method is proposed in a central air conditioning system of a university hall. A linear regression method is used to build the predictive power model and neural network (NN) algorithm is used to build the predictive temperature model. The two derived predictive models are then optimized with a multi-objective particle swarm optimization (MOPSO) algorithm. A detail study of the relationship between energy consumption and indoor comfort considering temperature, humidity, air handling units (AHU) and chiller machine is presented. Energy consumption and indoor comfort are realized by applying the control settings derived from the optimization model. A range of optimal control setting solutions for the central air conditioning system is derived by the MOPSO algorithm and provide a large set of trade-offs between indoor comfort and energy consumption.