Data Completion for Power Load Analysis Considering the Low-rank Property

With large-scale applications, the loss of power load data during transmission is inevitable. This paper proposes a data completion method considering the low rank property of the data. According to the low-rank property of data and numerical experiments, we find either the linear interpolation (LI) or the singular value decomposition (SVD) based method is superior to other methods depending on the smoothness of the data. We construct an index to measure the smoothness of data, and propose the SVDLI algorithm which adaptively selects different algorithms for data completion according to the index. Numerical simulations show that irrespective of the smoothness of data, the data complementing results of SVDLI are comparable to or better than the best of SVD or LI algorithms. The present study is verified using the measurements in China, and the public data of the Australian electricity distribution company and Lawrence Berkeley National Laboratory.