Stabilized Prediction Error Method for Closed-loop Identification of Unstable Systems

摘要

The prediction error method (PEM) based on output error is a versatile tool for tuning parameters in models from the I/O data of the target systems. However, if the target system is unstable and data is acquired in a closed-loop environment, a naive application of the PEM fails. For this problem, a stabilized version of PEM, where a virtual controller stabilizes the prediction error, is introduced in this paper. For linear models, the performance of the proposed method compares favorably with conventional closed-loop identification methods. While it has the versatility to accommodate a wide range of nonlinear models as long as it can simulate the output of the target system. In this paper, the performance and versatility of the proposed method are demonstrated through analysis based on linear models and numerical examples with nonlinear models.