Research On Air-Conditioning System Control Strategy Based On Artificial Neural Network

Module construction and feature extraction in traditional control mode can be avoided by adopting artificial neural network technology. Adverse effect caused by improper choosing of model and features can be eliminated, too. The neural network has strongly nonlinearity to shine upon ability, study adaptive captive. With its powerful fault-tolerant ability, users can simulate the control system without understanding the dynamic relationships (such as differential equation, transfer function, equation of state, etc.) between input and output variables of the object. Artificial neural network has obtained compelling achievements in load prediction, system identification, fault diagnosis, nonlinear optimization control, and other fields of HVAC system. It is more and more widely applied in air-conditioning system.Firstly, domestic and foreign current situations of the controlling system of air-conditioner are investigated and studied, and automatic controlling theory is reviewed in this paper. While analyzing the characteristics of multi-interferences, multi-operating modes, and coherence of temperature and humidity of air-conditioning system, conduct a series of studies to air-conditioning system in the market, and sets up building model, mathematical model and simulation model, and so on. Air-conditioning system is simulated on the basis of simulation model, two-position control and PID controlling mode, and its control effects are analyzed.Secondly, Analysis and study are carried on against BP neural network and BRF neural network, control modes are established and simulations are conducted for BP neural network and BRF neural network. Performances evaluation of these control modes is analyzed.At last, based on all of that, the application of combined control modes of artificial neural network and PID in air-conditioning system is studied. At the same time, build the simulation models of air-conditioning control system base on the MATLAB/Simulink artificial neural network toolbox. The weight of the BP network are conditioned, and then the PID Parameter on - line adjustment is to be achieved. The simulation results show that artificial BP neural network control and PID control system is more stability and overshoots smaller than compress BP neural network control or PID control. Results of Performance Evaluation show that: for the HVAC control systems, this control mode having very good control effect.