Characterization of Multimachine System Behavior Using a Hierarchical Trajectory Cluster Analysis

In this paper, an online hierarchical clustering method based on pattern recognition techniques is proposed for the automated clustering of system motion trajectories. Using the concept of minimum average distance between machine oscillations exhibiting a common behavior, a hierarchical clustered structure of the system that can be used for online determination of multimachine dynamic equivalents is suggested. The proposed formulation accounts for complex intermachine oscillations and is suitable for a wide range of problems such as wide-area stability analysis and online dynamic security assessment and control. The technique developed is applied to a six-area 377-machine model of the Mexican interconnected system. In particular, the clustering procedure is applied to identify the coherent motion of system machines following critical contingencies. Numerical examples illustrate the method and demonstrate the ability of the clustering technique to isolate and extract temporal modal behavior. It is shown that automated trajectory segmentation correctly identifies system dynamic behavior, enabling coherency analysis to be performed on an automated fashion. The method is currently being extended to perform model reduction, directly in time-domain using a multimachine representation.