新型电力系统负荷预测技术综述及基于数据空间的架构展望OA

In response to the"dual carbon"goals and the development requirements of novel power systems,addresses the limitations of existing research—namely,challenges in acquiring and managing massive,multi-source heterogeneous data,the limited scenario adaptability of prediction models,and the inadequate representation of global features using local ones.To this end,we present a comprehensive review of current load forecasting technologies and propose a prospective architecture,alongside an intelligent,optimized-selection forecasting mode for distributed,collaborative,and refined power load prediction based on the data space.First,existing power load forecasting studies are systematically reviewed and comparatively analyzed across four dimensions:influencing factors,forecasting scenarios,model algorithms,and performance evaluation metrics,thereby clarifying the current challenges in the field.Second,the concept of the data space is outlined,elucidating the mechanisms by which it empowers load forecasting.Finally,the proposed data space-based forecasting mode and prospective architecture are detailed through five operational stages(grid division,model library construction,method optimization,computational task allocation,and forecasting result fusion)and three functional modules(data preparation,data management,and data services).Ultimately,this study provides theoretical references and practical solutions for load forecasting within the complex spatiotemporal environments of novel power systems.