A Discrete Predictor Controller Applied To Sinusoidal Perturbation Adaptive Optimization

摘要

Uncontrollable and immeasurable disturbances possessing non-stationary characteristics often create major control problems in chemical processes. Typical of these fluctuations are composition changes in a homogeneous system and catalyst activity changes in a heterogeneous system.nIn this report, Box and Jenkins' theory of discrete predictive con¬trol is mechanized on an analog computer and applied to sinusoidal per¬turbation adaptive optimization of a natural gas combustion system. Process dynamics and the stochastic characteristics of disturbances are programmed into the discrete predictor controller to provide an adaptive control system which allows for process upsets before they occur. Statistically designed experiments are constructed to evaluate the performance.