Decomposition-based Adaptive Cross-modal Aggregation Algorithm for Heterogeneous Data in Safety Monitoring of Power Grid

The security warning of the power grid network is of great significance for the safe operation of the power grid. However, the monitoring data of the power grid is numerous and heterogeneous. How to effectively utilize these massive and heterogeneous monitoring data and extract effective warning information from them is currently a hot topic. To meet the above requirements, a decomposition-based adaptive cross-modal aggregation algorithm (DACMA) for heterogeneous data was proposed. In this algorithm, the view-invariant subspaces were firstly found to obtain transformed source and target domains based on data decomposition. Then, the joint learning was used to unify various types of monitoring data into a single model. At last, by adaptively adjusting, an adaptive approach for various types of monitoring data was constructed to provide a more comprehensive description of network threats. The experimental results showed that the DACMA algorithm had the efficient convergence and better prediction accuracy, which indicated that it had broad prospects in the field of safety warning in the power grid.