| Activated carbon(coke),as a recyclable and reusable adsorbent,has attracted more and more attention in recent years in the application of flue gas desulfurization and desulfurization.Its core part is the adsorption and analysis of combined pollutants and dust particles in flue gas by activated carbon.The control process of the adsorption tower has typical non-linear,timevarying and large inertia characteristics.The traditional PID control algorithm is difficult to achieve the desired control effect.Therefore,it is of great significance to study the application of intelligent algorithm in the control system of dry combined desulfurization and desulfurization of activated carbon.Based on the analysis of the coupling characteristics of activated carbon desulfurization and desulfurization adsorption tower,a decoupling control method using neural network PID is proposed.The effectiveness of the method is verified by simulation analysis.In this paper,the process flow,system structure and purification mechanism of dry desulfurization and desulfurization with activated carbon are thoroughly understood.Through the analysis of variable parameters in the process,the main disturbance variables and internal coupling relations in the control of adsorption tower are determined.Based on historical data,identification model is established by recursive least square method.Then on this basis,the control algorithm is designed.Firstly,the conventional PID algorithm is used to simulate the single closed-loop mode control.The control structure is separated into two separate control loops,and the PID parameters are optimized by using the differential evolution algorithm.According to the identification model,the feed-forward decoupling link is designed.The control curve with decoupling link and without decoupling link is analyzed when using conventional PID control,and the limitation of conventional control algorithm in desulfurization and denitrification control is analyzed.Then a self-correcting PID multivariable decoupling algorithm and a neuron-PID multivariable control method using RBF on-line identification are proposed.The self-correcting PID algorithm based on RBF model identification consists of a single neuron with self-learning and self-adapting ability to form the basic adaptive controller NNC.At the same time,the dynamic RBF neural network based on the least mean square function(LMS)of predictive output error is used to identify the object model online and improve the adaptive ability of the controller.It is extended to the multivariable control mode,and the multivariable decoupling control of the adsorption tower is completed by the self-learning of the neural network.The controller using PID neuron for multivariable control is simple in structure,easy to implement,does not need specific decoupling and identification links,and has good adaptive ability.Finally,the hardware network structure of the control system is described,and the selfcorrecting PID decoupling algorithm based on RBF model identification is programmed by structured text(SCL)language,which is a PLC programming language conforming to IEC61131-3 specification.The control effect of the algorithm and the integrity of the algorithm function block are verified by simulation,which proves that the algorithm is effective and implementable.By encapsulating the algorithm code as FB function block or library template file,it can be called many times,and it can be easily transplanted and applied on different brands of PLC,PAC or DCS software and hardware systems conforming to IEC61131-3 specification.The algorithm designed in this paper has high application and popularization value. |
PDF Full Text Request