Parametric identification of state-space dynamic systems: A time-domain perspective

In this paper we have presented a time-domain approach to parametric identification of state-space dynamic models comprised both an equation of motion and a system potential (a performance measure). The proposed techniques have been elaborated in order to obtain high simulation and forecast properties and applied to systems of nonstationary accelerator, gradient systems, and linear-quadratic stationary systems. We have also demonstrated a new concept of system potential specification in case of linear-quadratic stationary systems. It is based on the principle of its basis decomposition as an element of energy space. All models and algorithms have been approbated using real statistical data for models of macroeconomic dynamics.