Predictive control using an adaptive fast model

The performance of predictive control systems is dependent on the accuracy with which the fast model represents the plant. When the parameters of the plant vary appreciably with time, the use of a fixed-parameter model results in nontime-optimal responses. However, by employing an adaptive scheme in which the parameters of the plant are continuously estimated and the parameters of the fast model are continuously updated, the response remains time optimal. The paper is concerned with a particular scheme of this type for a variable-parameter second-order plant in which one of the time constants is dominant. The parameter estimator employed measures the dominant time constant continuously during transients of the system and measures the gain coefficient in the steady state. In addition, changes in the gain coefficient during transients are taken account of to a first order by a modification of the estimated time constant. The parameters of the model are thus correct when most needed; i.e. at the instant of predicted switching. The control scheme, which incorporates a small linear band to prevent limit cycling, was developed for the voltage control of an isolated synchronous generator, and experimental results are given from a 3kVA set. In addition, results are given for the control of various simulated plant showing the performance of the scheme for time-variable-parameter and higher-order plant.