Aggregated Load Uncertainty Analysis Through Unsupervised Learning

Climate changes in recent years have raised serious concerns for governmental, academic and industrial communities. One critical mitigation strategy identified was to reduce the carbon emission. New power systems thus have been strategically proposed to achieve the carbon neutrality target in China by 2060. It requires the coordinated operations among the generation, network, load and energy storage sectors, where the transmission and distribution networks are also more tightly connected. Load analysis, specifically load uncertainty analysis, would be necessary to provide deeper insights for the system operation and planning. In this paper, we proposed an unsupervised learning based aggregated load uncertainty analysis approach. Tensor decomposition was applied to decompose the aggregated load profiles collected through SCADA, PMU, smart meters or distribution-level PMUs. The joint aggregation effects were analyzed through the decomposed component profiles. Uncertainty information was extracted through such profiles accordingly. Numerical results obtained from real load data demonstrate the effectiveness of the proposed approach.