Predictive Control on Bed Temperature of CFB Boilers Based on Multiple Models

Aiming at the strong nonlinearity features of the bed temperature of circulating fluidized bed(CFB) boilers,a modified predictive control scheme based on multiple models has been proposed,which combines multi-model control with generalized predictive control,and uses models of several typical operating points to approach the dynamic characteristics of the total range of operating points,so as to correct models instead of on-line identification.By tracing the change of actual working conditions for weighting of local controllers,appropriate control increment is obtained,while the object function of generalized predictive control modified so that the oscillation at later stages can be eliminated.The models of typical operating points are identified off line,of which the parameters are generally constant,so that solution of Diophantine equation and calculation of control law can be done off-line,resulting in reduced amount of on-line calculation.The algorithm was applied to the bed temperature control of a 440 t/h CFB boiler.Simulation results demonstrate that the control method may effectively overcome the nonlinear shortcoming of the controlled object,via which better control effect may be obtained even under large variations of load.