An Intelligent Multi-Source Data Clustering and Analysis Method for AC/DC Hybrid Power Grid Planning with Increased Complexity

In order to better analyze the response characteristics of an AC/DC hybrid power grid with increased complexity following various types of disturbances, this paper proposes an intelligent method for clustering dynamic responses of multi-source data obtained from actual measurements or high-fidelity simulators. Base cases for AC/DC hybrid power grid planning students are created first, and a list of high-impact contingencies is searched using a graph algorithm. Then, the electromechanical-electromagnetic hybrid simulation method is adopted to create multi-source simulation data in terms of dynamic voltage, frequency, etc. An unsupervised learning algorithm and cluster analysis are then conducted to extract key features of dynamic response characteristics of the AC/DC hybrid power grid and create specific cluster labels. Advanced analysis, e.g., fast evaluation of risks from multi-source steady-state data, can then be enabled for further analysis. The proposed approach's effectiveness is verified on the actual AC/DC power grid model in East China.