Multi-granularity Signal Processing Method for Digital Twin Power Grids via Graph Representation Learning

This study introduces an innovative signal processing approach employing a multigranularity aggregation method grounded in graph representation learning, addressing the need for comprehensive attribute analysis in real-time modeling of digital twin power grids. Traditional algorithms often focus narrowly on isolated node information, inadequately capturing the holistic characteristics of information networks. By integrating signal processing techniques, this method enhances the incorporation of node interdependencies, accounting for both spatial distances and business attributes within the network topology. This technique seamlessly merges topological and business data across various informational layers, enabling multi -granularity clustering and mapping at the unit, system, and complex system levels within digital twin grids. The proposed approach significantly advances the application of signal processing in the dynamic analysis and decision -making processes essential for optimizing digital twin grid operations.