Big-Data Multimodal Power System Analysis with Markov Temporal Modeling for Community Detection

Abstract

Modern power systems generate massive volumes of heterogeneous data from sensors, meters, supervisory devices, market platforms, and environmental services. Extracting stable structural patterns—especially community structures—from such big-data environments is essential for understanding grid behavior, enhancing situational awareness, and supporting reliable operation. However, existing community detection approaches often fail to capture the non-stationary temporal dependencies and multimodal heterogeneity inherent in large-scale power networks.

To address these challenges, this study proposes a big-data multimodal learning framework for power system community detection, integrating electrical measurements, topological descriptors, environmental indicators, and operational event logs into a unified representation. A Markov temporal prior is employed to model sequential dependencies and suppress unstable transitions across evolving power system states, thereby improving temporal consistency. The framework further incorporates scalable data processing, adaptive modality fusion, and topology-aware clustering techniques tailored for large and complex power grids.

Experiments conducted on both real-world and synthetic datasets demonstrate that the proposed framework achieves superior accuracy, robustness, and temporal stability compared with state-of-the-art baselines. The detected community structures align closely with physical grid regions, operational partitions, and disturbance propagation patterns, confirming the method’s practical interpretability.

This work provides a scalable and data-driven solution for community detection in large power systems, offering valuable insights for grid monitoring, anomaly localization, reliability assessment, and intelligent decision-making.