Novel generalized predictive control algorithm with application

A novel generalized predictive control algorithm with offline and online identification is presented in this paper. Unlike the existing GPC algorithms, the proposed approach decomposes the model parameters into a time-invariant part and a time-varying part and offline identification and online identification algorithm is used respectively to treat both of them. Therefore the reliability and precision of the predictive model are are improved simultaneously. A simulation example of FBR (fixed bed reactor) plant shows that this new algorithm is not only reliable in the case of plant uncertainties and outside abnormal disturbances, but is also adaptable for slow time varying processes and can effectively overcome the disadvantageous influences of modeling errors.