User electricity stealing behavior identification method based on high-dimensional random matrix

Abstract

The invention discloses a user electricity stealing behavior identification method based on a high-dimensional random matrix. The high-dimensional random matrix is constructed by collecting the metering parameters acquired by the user terminal electric meter, and the user electricity stealing behavior is identified by analyzing the distribution relationship of the high-dimensional random matrix covariance matrix feature value and the high-dimensional random matrix covariance matrix feature value function limit convergence function trajectories on the complex plane; and the mean spectral radiuschange curve is further solved for the user having the electricity stealing behavior, and the start and end time of user electricity stealing is located through the fluctuation situation of the meanspectral radius change curve and the electricity stealing duration is calculated. Larger data scale calculation can be borne, the multidimensional relation of time and space of multi-parameter auto-correlation and cross-correlation can be comprehensively considered, the identification result is more accurate and the leak detection and false detection rate is lower. The method also has the capacityof locating the start and end time of electricity stealing which is not possessed by the conventional identification method so that the method has great extendibility and can adapt to the future development demand.

Description

technical field [0001] The present invention relates to the field of power metering and inspection, and more specifically, to a method for identifying user electricity theft behavior based on a high-dimensional random matrix. Background technique [0002] Stealing electricity by interfering or transforming metering devices will cause non-technical losses to the power grid and cause economic losses to power companies. The traditional identification of electricity stealing behavior mainly relies on regular inspections by power inspectors, emphasizing manual management, which requires a lot of manpower, financial and material resources, and cannot achieve the expected results. With the popularization of smart meters and the rapid development of power system informatization, power companies have stored massive user-side power data. By fully exploiting the potential value of electric power big data, timely identifying user stealing behavior and reducing economic losses are of gr...

AI Technical Summary

Problems solved by technology

For example, units such as New York University and Chongqing University of Posts and Telecommunications use clustering algorithms to analyze load curves to identify user behaviors; users; Shanghai Jiaotong University has successively used neural network, time series analysis and other methods to analyze user electricity theft behavior; the above research has achieved certain results in the identification of user abnormal electricity consumption behavior, but these algorithms are often only for small-scale data and single parameter data However, the metering system is a multi-parameter synchronous measurement system. The impact of stealing power on the system is often manifested as a synchronous comprehensive change of multiple parameters, and different power stealing methods are manifested as a synthesis of different parameters. Therefore, traditional methods only use The analysis and identification of a single parameter data source is often prone to missed detection; in addition, when the user load changes, it will also cause changes in some measurement parameters. If only a single parameter is used for analysis and detection, false detections will also occur.

More importantly, traditional methods can only identify whether a user is stealing electricity, but are powerless to determine when to start stealing electricity and how long the electricity stealing behavior lasts. Traditional algorithms are in terms of computing scale, functional requirements, accuracy, robustness, Inadequacies in reliability and efficiency are obvious

Images