Control Structure Selection Of Decentralized Pid Control For Hammerstein Nonlinear System

In the actual industrial process control,the controlled object is generally a multi-variable nonlinear system,which has a certain degree of coupling and whose input to output mapping is related to the working point.Considering the control requirements for safety and reliability,the bottom control system of industrial process mainly adopts decentralized conventional PID control,but it is difficult to apply PID control to multivariable nonlinear systems.In this paper,the Hammerstein multivariable nonlinear system is studied,which consists of a nonlinear steady-state model ahead and a linear dynamic model back.The artificial neural network and describing function are used to deal with nonlinear relative gain array respectively,and the control loop configuration for Hammerstein multi-variable nonlinear system is achieved.In the time domain,the interaction analysis method is designed based on energy gain and artificial neural network.The linear dynamic model is transformed into a steady-state energy gain array,and a nonlinear steady-state gain array is produced through combining the nonlinear steady-state model and the linear steady-state energy gain array.Then two artificial neural networks are used to fit the forward mapping and reverse mapping of the combined nonlinear steady-state gain array.Using the trained forward and reverse artificial neural networks,the nonlinear relative gain array is obtained for control loop configuration.In the frequency domain,the interaction analysis method is designed based on describing function and frequency dependent dynamic gain.The description function of nonlinear steady-state model is multiplied by the open-loop frequency response of linear dynamic model,and a frequency dependent relative gain array is obtained which is also dependent on input amplitude.Then,averaging the distances between the elements of the array and(1,j0)in complex plane,the average coupling coefficient matrix is calculated,and the control loops are built for the input and output pairings corresponding to the minimum coupling coefficients on each row.When the nonlinear steady-state model is decoupled,the relation between PI controller parameters and stability margins is deduced for the paired control loops.So,the feasible region of PI controller parameters can be found to obtain the appropriate PI controller parameters.