Research On Modeling And Optimal Control For Central Air-conditioning System Based On Time Delay Identification

The theoretical model,empirical model and data model of central air-conditioning system,which can reflect the thermodynamic characteristics and determine the control performance,are the research hotspot.Where,Theoretical model is an engineering model using mathematical logic and language,and the common method to reflect the internal parameter change of the modeling link is the energy balance.Empirical model simplifies the coefficients which are difficult to be determined or large variation range in the theoretical model according to the experience of researchers.Ddata model is an abstract representation for the data characteristics of the modeling link with the historical data as the analysis object.While,the central air-conditioning system is with the performance of time-delay which influence the changes of room temperature and thermal state,and this phenomenon will be considered in any models.For describing the time-delay phenomena,it is usually to add the time-delay when the basic model is established or add the time-delay value in the time variable,and there are not many researches on the determination of time delay.As a basic problem research,it has a positive role in the optimization of system modeling combing with more reasonable analysis of time-delay phenomenon.Moreover,the model is also beneficial to the control optimization in the system.The control scheme based on model can be divided into PID control,fuzzy control,neural network control,predictive and preview control control in central air-conditioning system.The action time of the controller and the actuator will be reduced when the model can reflect and predict the time-delay stability zone or time-delay value in the process of parameter adjustment,and the system state parameters will reach the standard quickly with saving energy.There are mass data in the building energy management system,and the the value of time-delay cannot be obtained directly due to the long data sampling period of central air-conditioning system.For providing the condition for the optimization of system modeling and control methods,the time-delay identification needs to combine many data processing methods to obtain good historical data quality and transform the data representation.In this paper,the time-delay modeling and identification of the central air-conditioning system are researched to solve the above problems,and time-delay state-space model and data model are designed.The details are as follows:First of all,the definition of time-delay is presented to distinguish the similar character and determine the scope of time delay in the central air-conditioning system.For time delay,the influence about system operation and control is analyzed,data presentation and time delay compensation scheme are explained in the case building.Meanwhile,the time delay identification based on data-changing trend and simulated annealing algorithm is designed using the optimized data which is disposed by particle swarm optimization and Kalman filtering.Secondly,linear and nonlinear model are designed to predict future parameters about thermal state and time delay in central air-conditioning system.In the linear modeling,time delay state space model is used to embody the time delay phenomenon and calculate the time delay stability interval based on the energy balance,and it is applied to valve optimized control combining with the preview control.In the nolinear modeling,modified Elman and BP neural network which chose the trend change data as the connection and constraint conditions between them are used to predict the time delay and state data for the modeling part.According to the acceptable results,an adaptive control scheme combining with Fuzzy-PID control(Fuzzy-PID-M)is design for the application of this nolinear model.Comparing with traditional Fuzzy-PID control,more precise control performance can be abtained using the Fuzzy-PID-M,and the response delay,settling time and energy consumption are decreased.Finally,the performance and application condition of linear and nonlinear model are compared in order to select the modeling approach and control scheme for different building and data conditions.