Characteristics Parameters of Electromechanical Oscillation Quick Identification Method under Ambient Excitation based on Bayesian DMD

The measurement response data contain a vast amount of dynamic information about the system. Due to the widespread use of wide-area measurement systems, methods for identifying oscillation characteristic parameters driven by random response data have found wide applications. Building on the characteristic analysis of random response data in power systems, the Bayesian Dynamic Mode decomposition algorithm (Bayesian-DMD) is proposed. The main idea behind the Bayesian-DMD method is to incorporate Bayesian principles into DMD, establishing a probability model corresponding to DMD, and then providing a Gibbs sampler for Bayesian DMD’s posterior inference. This is applied to complex matrices comprising generator angular frequencies, generator angular velocities, and active power combinations, thereby reducing the dimensionality of the dataset and improving the accuracy of eigenvalue calculations. Eigenvalues can be used to compute modal frequencies and damping ratios, and the Bayesian-DMD algorithm can determine modal shapes. From the perspective of the measurement data matrix, this approach enhances the computational speed and robustness of dynamic mode decomposition. This paper introduces the theoretical background and process of the Bayesian-DMD algorithm, conducts simulation tests on real measurement random response data from a foreign actual system with the IEEE 16-Generator 68-Bus power system, and compares it with the Stochastic Subspace Identification (SSI) method to validate the algorithm's effectiveness.