A State Sensing Method Based on Multi-Source Data Fusion for Virtual Power Plants

Due to the deep coupling of modern information technology and power-gas-heat diversified energy networks, the detection of changes in operating characteristics, such as wind turbine off-grid events, has become an urgent problem that needs to be addressed, which is essential to ensure the safe operation of virtual power plants. In the context of three-innovation integration, where multiple types of intelligent terminals are deployed, large-scale system nodes are interconnected, and the operating characteristics are both complex and variable, an event detection method based on the random matrix spectral distribution is introduced. First, a standard covariance matrix model is established to fuse heterogeneous data from multiple sources, reflecting the coupling relationship of different energy subsystems. Second, a relative entropy-based spectral distribution difference measure is proposed to detect changes in system oper-ating characteristics, and the nodes responsible for these changes are localized using the maximum eigenvectors corresponding to the spectral distribution. Finally, the effectiveness of the proposed method is validated through simulation analysis of various event types.