On-Line Identification Simulation of Forgetting Methods to Track Time Varying Parameters Using Alternative Covariance Matrix

The paper compares abilities of forgetting methods to track time varying parameters of two different simulated models with different types of excitation. The observed parameters in simulations are the integral sum of the Euclidean norm of a deviation of the parameter estimates from their true values and a selected band prediction error count. As supplementary information we observe the eigenvalues of the covariance matrix. In the paper we used modified method of Regularized Exponential Forgetting with Alternative Covariance Matrix (REFACM or REZAKM) along with Directional Forgetting (DF or SZ) and three regularized methods has been selected: Regularised Exponential Forgetting (REF or REZ), Stabilised Linear Forgetting (SLF or SLZ) and Praly Forgetting (PF or PZ).