Data-driven Event-triggered Prediction Control For Density Loop Of Dense Medium Separation Process

Dense medium separation technology is one of the main technologies of coal washing,which plays an important role in optimization of energy structure in our country.The density loop of dense medium separation adjusts the density of dense medium suspension by adding water and medium to track the set points,so as to ensure the efficient operation of dense medium separation process.With the development of the industrial internet,the intelligent control algorithm has been closely integrated with the coal preparation process,which has important economic and environmental protection value.In the density loop of dense medium separation,the detection of dense medium suspension density is limited by hardware,which extends the sampling period of the output.In order to achieve better performance under transient conditions,the opening period of the water valve is usually set to be less than the output sampling period.Therefore,there is a multi-rate problem in the dense medium separation density loop.Secondly,equipment aging and other conditions inevitably exist in dense medium separation density loop,which leads to the slow time-varying characteristics of the model parameters.In addition,with the development of the industrial internet,the information transmitted in the dense medium separation system is no longer limited to data,but also includes images,videos and other information,which leads to a significant increase in the amount of information transmitted in the system,and even channel congestion in serious cases,affecting the normal operation of the separation system.Therefore,the data-driven event-triggered predictive control of the dense medium separation density loop was carried out to study the multi-rate,slow time-varying parameters and channel congestion of the dense medium separation density loop respectively.The main research work is as follows:(1)The process and mechanism model of the dense medium separation density loop are deeply studied,and the control problems such as multi-rate,slow time-varying parameters and channel congestion are further analyzed in detail.(2)A multi-rate density loop control method based on event-trigged is proposed.Firstly,the lifting technology is used to unify the different time scales of input and output.Furthermore,a multi-rate parameter identification method based on continual learning is designed to identify the model parameters online.Secondly,a multi-rate event-triggered model predictive controller is designed to reduce the times of controller updates while ensuring control performance.(3)A data-driven multi-rate density loop control method is further proposed.This method directly uses the input and output data of different rates to directly design the multi-rate controller,and introduces the concept of rolling horizon to predict the data.On this basis,the correction term is obtained by solving the optimization problem,and the control law can be obtained by combining the correction term with the prediction process to realize the tracking of dense medium suspension density.Secondly,the event-triggered mechanism is introduced to make the controller update nonperiodically to reduce the waste of communication resources.The thesis includes 21 figures,1 tables and 64 references.