On-line order selection and parameter estimation--An experimental application

The experimental application of an integrated on-line order determination and a parameter estimation algorithm to a storage plant is presented. The experiments are carried out on different operating points around which the plant can be modeled as a linear, discrete-time, stochastic and S.I.S.O. system. A simple algorithm for the determination of the order is proposed. The experiment is based on testing recursively the significance of the reduction in the prediction error variance when increasing the model order, by means of a modified F-test. The results of three well-known identification procedures, used to perform the parameter estimation in different conditions, are also reported for comparison purposes.