Optimal PSS-Parameter Selection Algorithm with New Performance Measure

Abstract

In this paper a parameter optimization method with a new performance measure is proposed for the selection of PSS parameters. Outputs (or states) of a linear or linearized system are composed of various modes. So it is possible to regulate outputs (or states) by suppressing all modes or concerned modes in outputs. Thus here the performance is defined as weighted sum of areas under squares of concerned mode envelopes. Since all the modes are represented by eigenvalues and eigenvectors, they are affected by the parameters contained in system matrix. That is a reason to adopt a parameter optimization approach. The steepest descent method is used for minimizing the performance. The gradient of the performance with respect to PSS parameters is derived in an explicit form. In the optimization procedure it is required to calculate eigenvalues and eigenvectors at each iteration of updating parameters. Its computation time is quite a burden but can be reduced by a sensitivity analysis. Eigenvalue and eigenvector sensitivities with respect to parameters are applied to update them when parameters are changed to better directions. The proposed algorithm is applied to a one-machine infinite-bus system and two machine system. Simulation results show improvement in the stability of sample systems.