Research On Big Data Compression Method Of Smart Power Distribution Systems Based Tensor Decomposition

Data compression of power distribution system is an important topic in the field of power system.However,the power distribution system still lacks simple and effective homogeneous and lossless representation of structured and unstructured data.Because traditional data compression methods require matrix(or vector)representation of data,the high order dependence of data is destroyed and the compression precision of power distribution system data is seriously affected.In addition,the current static data compression methods do not take into account the feature of big data accumulation over time in power distribution system,resulting in low processing efficiency during compression.Based on the three problems of power distribution system data compression,this paper studies the homogeneous representation method of all types of distribution system data based on tensors,and proposes a high-precision power distribution system tensor Tucker decomposition data compression method and a high-efficiency power distribution system incremental tensor decomposition data compression method.(1)Based on tensor theory,this paper constructs various data representation models of power distribution system that are concise,homogeneous and lossless,and realizes the concise,homogeneous and lossless representation of structured operation data and unstructured state monitoring(image and video)data of power distribution system.It provides basic guarantee for the high precision compression and processing of big data in power distribution system.(2)Based on the data representation method of power distribution system constructed in(1),this paper proposes a power distribution system big data compression method based on tensor Tucker decompression.Compared with the traditional matrix(or vector)compression method,this method can preserve the high order dependence of data during compression.The real power distribution system data is used to verify that the proposed power distribution system data compression method based on tensor Tucker decomposition can effectively reduce the power distribution system data volume.And,the comparison with the singular value decomposition method shows that the proposed method is much more accurate than the singular value decomposition method.(3)Based on the feature of big data accumulation over time in distributio power system,this paper proposes the method of power distribution system big data incremental tensor decompression combined with incremental update technology.In the process of compression,the method adopts the technology of mode expansion and incremental update.Therefore,the high dimensional spatial characteristic structure of power distribution system data is maintained to ensure high compression accuracy,and the time and memory consumption of compression method are reduced.The actual verification results in a power distribution system industrial park shows that this method can effectively compress the data of the power distribution system and has the advantages of high precision,short time and small memory consumption.