Research on Identification Process of Nonlinear System Based on An Improved Recursive Least Squares Algorithm

The global dynamics for nonlinear system commonly exist in practical application is built through multiple local linear models, however, least squares method is only suitable for the latter, solving the problem by a robust recursive least-squares method to identify the system. Common parameter identification algorithm is only appropriate for slow time-varying systems, but the proposed improved algorithm is effective for condition that the parameters change rapidly and difficult to track in real time. It shows that the improved least squares algorithm can extend parameter estimation range. Recursive least square method can obtain parameter estimations of noise model and process model simultaneously, however, traditional least square method can only realize parameter estimations of process model. The convergence performance of the raised algorithm will be demonstrated. Simulations are given to illustrate the availability and correctness of the proposed method.