A new parameter estimation algorithm for time-varying systems and its application to a cement blending process

A new parameter estimation method for time-varying systems is proposed in this paper. First, by introducing ´a priori´ knowledge into the performance index of least square (LS) algorithm, a modified LS algorithm is developed. Then, a number of modifications have been proposed to improve some existing LS algorithms, such as the real-time recursive LS algorithm and the instrumental variable (IV) algorithm. Finally the newly proposed method is applied to a cement blending process. The test results show that the new LS algorithm has remarkably improved the estimation accuracy and the convergent rate of tracking to the true time-varying system parameters.