Improvement of parameter estimation using prediction error dynamics for unstable system

In this paper a parameter estimation of an unstable system is presented. In general ARX or ARMAX model structure to be identified results stable predictor. Even though the model structure has generically stable predictor, we have no such method that can validate the model structure and results of parameter estimation. In the estimation of parameter for stable system, one validates the estimation results by output prediction. But in the unstable system, although we have parameters to be estimated very accurately, model validation by output prediction is very poor. This paper proposes the method using the dynamics of prediction error for this problem to identify unstable system parameter. We apply this method to a real system to be unstable to estimate the parameters of unstable system.