Aiming at the potentially associated multi-group features and high-dimensional multi-source heterogeneous data in the power grid, this paper utilizes distributed technology to address data updating between multi-level heterogeneous spatial databases. The tasks of the centers at all levels are to receive, store, process, manage and apply various types of resource data needed for resource management and social services, and to realize synchronous or asynchronous updates with remote data from higher-level data centers and lower-level data centers through the Resource Information Network (RIN). Through the distributed parallel framework and deployment mapping and subsumption services, the analysis and calculation process of power marketing data is established, and the data processing workload of the headquarter-provincial company is assigned to the high-performance computing cluster for execution, and its calculation results are stored in a distributed manner so as to facilitate redistributed calculations after changing the algorithms and parameter settings of power marketing data. The sparse modeling realized by LASSO regression can effectively reduce the computation time, and can support the steady progress of the online assessment of voltage stability margin. The experimental results show that the load forecasting error rate of this paper’s architecture is 3.5%, the accuracy of customer segmentation is 91%, and the average acceleration ratio is 3.83, and the framework can provide guidance and support for the stable operation of the power system.
