A Big Data Architecture Design for Smart Grids Based on Random Matrix Theory

Abstract: Data in smart grids with features of volume, velocity, variety, and veracity (i.e. 4Vs data) are difficult to handle by traditional tools, which are highly dependent on assumptions of specific roles or models with casual logics. This paper motivates big data analysis to process data from high-dimensional perspectives by using random matrix theory. An architecture combining smart grids and big data is proposed as a universal solution for control and operation in power systems. Based on this architecture, we are able to detect signals indicating sudden changes or faults in a power grid by comparing experimental findings with the random matrix theoretical predictions. Mean Spectral energy Radius (MSR) is defined as a new statistic to visualize the data correlations for this architecture. Comparative analysis of the MSRs from distributed regional centers under group-work mode is able to produce a contour line to locate the signal source even with data of imperceptible differences in low-dimensional perspectives. It demonstrates that some analyses are able to be extracted directly from the raw data which was hardly observable by conventional tools. Five case studies and their visualizations validate the effectiveness and higher performance of this designed architecture in various fields of smart grids. To our best knowledge, this study is the first attempt to design such a universal architecture for applying big data in smart grids.