Time series identification of interarea and local generator resonant modes

A power system consisting of large generation pools is characterized by several modes of oscillation. The various modes arise from a group of generators swinging relative to other groups as well as each machine swinging relative to its local bus. A practical problem that arises in the design and evaluation of controllers for damping power system oscillations is to measure the appropriate transfer function(s) of the power system. The transfer function from the input to the power system to the output signal being fed back to the controller contains eigenvalues that characterize each of the modes embedded in the time-domain transients. The number of modes, the range of residues (or amplitudes of each mode), and the random disturbances that appear on the measured output signals, all conspire to make this a challenging problem. In this paper, the least squares and generalized least squares time-domain identification algorithms are applied to a simulated power system exhibiting local and interarea oscillatory modes. A comparison in the frequency domain is made between the simulated system and the models obtained by the two identification techniques. Under the experimental conditions used in this paper, both the local and interarea oscillatory modes are identified by the generalized least squares identification technique while the least squares parameter estimation method identifies only the local mode