| Improving the control system performance and operation efficiency of thermal power units is not only beneficial to ease the increasingly tight energy and resource constraints,but also beneficial to improve the peak and frequency modulation performance of thermal power units with a wide range of load changes.However,Proportion Integration Differentiation(PID)control performance is limited and it will take some time to replace PID with advanced control strategy.Therefore,it is urgent to improve the control performance of PID control system.The research foundation of the control system is the precise object model,and the thermal power units have accumulated a large number of operation data,which provides a data basis for the closed-loop identification based on the historical operation data.In this paper,the research of closed-loop identification modeling and control optimization is carried out for thermal power unit thermal work object.The main contents include:1.In order to obtain an accurate closed-loop identification model for thermal objects,a closed-loop identification process was proposed,which used historical operation data,integrated identification input variable selection,model order determination and unbiased parameter estimation.Variance expansion factor method and partial least squares regression method were used to select the input variables,and asymptotic method was used to identify the parameters.Taking the deaerator water level system as an example,the identification model was combined with the actual field control strategy to build a closed-loop simulation platform,and the reliability and accuracy of the above-mentioned identification process were verified.2.In order to ensure better identification results of historical operation data used for closed-loop identification,a method of identification data sharing and screening based on feature classification machine learning was proposed.First of all,a kind of data characteristic structure rules based on the historical operation data was proposed,and it can better express the historical operation data of dynamic information.In addition,the output of the closed-loop identification model fitting as whether historical operation data for closed-loop identification threshold to divide the data labels,so as to construct sample collection.Then,a closed-loop identification data classification rule model was established by using random forest and extreme forest algorithms to screen high quality historical data.The simulation results show that the classification prediction accuracy of the above identification data optimization method is high,and it can provide data support for the closed-loop identification based on historical data.3.In order to improve the control performance of the PID controller,a PID parameter optimization method combining the actual operation data of the thermal object,considering the output deviation,the total input variation,the maximum sensitivity of the system and the closed-loop stability was proposed.The simulation results show that the performance of the controller is greatly improved after the optimized parameters,and it is applied to the actual project to verify the effectiveness of the controller.4.In order to improve the control performance of thermal power units with PID as the main control method,a linear quadratic regulator optimal dynamic feedforward controller was proposed,which is not limited to the minimum phase object.Then,a disturbance compensation PID feedback control strategy based on generalized extended state observer was proposed.The disturbance compensation matrix of the generalized extended state observer is designed based on the linear quadratic regulator theory and combined with PID.Finally,the simulation is carried out in the deaerator water level control system.The simulation results show that the above control strategy improves the control performance of the PID system and is conducive to improve the control performance and economy of thermal power units. |
PDF Full Text Request