Main steam temperature multi-model prediction and control method based on a multi-model set

Concerning a kind of industrial processes for which first-order inertia plus a pure lagging model can be used to describe their dynamic characteristics under different operating conditions and which change with operating conditions, a method was presented for setting up a multi-model set based on the maximum and minimum values of the characteristic parameters of an object. A recursive Bayesian probability weighting method was used to obtain an overall predictive model. On this basis, a multi-model predictive controller was designed to meet the control requirement for the operating conditions varying in a wide range. In the meanwhile, when a rectification of errors is being performed, the prediction error of the model resulting from any dynamic change of the operating condition can be compensated in advance to enhance prediction accuracy. The simulation calculation results of a utility boiler main steam temperature system show that the method under discussion enjoys a superior ability to track a set value under various operating conditions. When the operating conditions change in a wide range, it is possible to stabilize the main steam temperature near a set value.