Adaptive Algorithm for Identification Parameters of Linear Nonstationary Systems

In this paper, we consider the problem of identifying unknown parameters of non-stationary systems. Let us assume that the non-stationary parameter of the system can be represented by the outputs of linear generators with an unknown state matrix and initial conditions vector. It is proposed that the state, control signal, and output variable are measurable. In the first step, the problem of parameterizing the initial dynamic model into a linear static regression model is solved. The second step is to estimate the unknown constant parameters of the linear regression model using the Dynamic Regressor Extension and Mixing method, which allows obtaining monotonic estimates and ensures the acceleration of the convergence of the estimates to the true values. The results of computer simulation showed the efficiency of the developed algorithm.