Abstract: The time -varying AR model can be applied to analyse nonstationary signals;the parameters estimation often uses the Least Sqnare methodology now. This papelr provides several problems in LS estimation of this kind of model,and an othogonal LS algorithm is proposed.High accuracy is achieved in predicting the control of the dynamics of pH in cyclic cooling water.

Key words: nonstationary, time-varying AR model, othogonal least square, ill-conditioned