System identification techniques for power systems analysis using distorted data

In this paper the performance of four system identification methods is evaluated for the analysis of different power system configurations. The methods considered are: nonlinear least squares (NLS), Prony method, sub-space state space system identification (N4SID) and the prediction error method (PEM). Artificially created data distorted by noise are used to represent real-world conditions. The analysis verifies the practical value of system identification methods for power system dynamic analysis and also illustrates practical issues and solutions encountered in their application.