ABSTRACT
Currently, the building of power grid middle platforms plays an increasingly important role in the development of the smart grid. However, due to the property of real-time sampling, middle platform data suffer from a series of abnormal problems. These contaminated incomplete data greatly hinder grid analysis accuracy and efficiency, and further cause catastrophes for the stable running of the smart grid. To address this challenge, in this paper, a novel abnormal value imputation method for smart power grid middle platform data is designed to deliver more accurate recovered results. Technically, original middle platform data are reconstructed to a high-order tensor form which delivered a more precise representation of time-continuous data. Then, the tensor-based abnormal value imputation model is formulated to a solvable convex function. Furthermore, the alternating direction method of multipliers is developed to optimize the proposed model. Finally, extensive experimental evaluations are conducted in the power grid middle platform dataset to elaborate the superiority of the proposed method by comparison with multiple state-of-the-art abnormal data imputation methods.
