An Investigation into the Application of Digital Prediction Filters to Direct Digital Control Systems

The report describes a digital simulation study of the application of discrete prediction filters, designed from Taylor series expansion or the method of least squares, to direct digital control, where the prediction filter is inserted in the feedback path. The effects on system performance to set-point changes and load disturbances are investigated. In addition, the effects of process noise and system stability are considered. It is concluded that accuracy of prediction is not the most important property which affects control performance. Best performance is obtained using a filter which has a low sensitivity to process noise and predicts as far as possible into the future while remaining within stability constraints.