Multi-dimensional intelligent analysis experience algorithm and classification algorithm for judging default electricity utilization and electricity stealing

Abstract

The invention discloses a multi-dimensional intelligent analysis experience algorithm and a classification algorithm for judging default electricity consumption and electricity stealing. The algorithm comprises the following steps: step 1, establishing a user tag library based on multi-source data fusion; step 2, establishing machine learning to construct a default electricity utilization and electricity stealing analysis model; and step 3, continuously training, perfecting and improving through investigation results and experience learning. Through training, verification and improvement of various models, the calculation efficiency of the models is optimized, and the accuracy, timeliness and reliability of model analysis are improved. And mutual verification and self-promotion of the processing flow of default electricity consumption and electricity stealing, an experience algorithm and a classification algorithm are realized. And if the abnormity is not eliminated, the system is automatically brought into training again, the analysis model is updated, and the self-learning and self-optimization functions are realized. According to the invention, judgment and positioning of default electricity utilization and electricity stealing become very convenient and simple. The method can be widely applied to judgment and accurate positioning of default electricity utilization and electricity stealing.

Description

technical field [0001] The invention relates to the judging and positioning technology of power violation and stealing electricity, belongs to the technical field of intelligent power utilization, and specifically relates to a multi-dimensional intelligent analysis empirical algorithm and classification algorithm for judging power violation and stealing electricity. Background technique [0002] At present, a small number of power users are driven by economic interests, and electricity theft occurs frequently. With the rapid development of science and technology, there are more and more acts of stealing electricity and using electricity in breach of contract through high-tech means. These technical means are not only highly concealed, but also difficult to investigate and deal with. Now the users of stealing electricity and using electricity in breach of contract have spread from specialized users to general industrial and commercial and low-voltage residents. Stealing ele...

AI Technical Summary

Problems solved by technology

Due to the intelligence, diversification and concealment of current electricity stealing methods, such as high-tech electricity stealing methods such as strong magnetic fields, radio frequency radiation, strong electric field interference, and attacking electric energy meters, usually leave no traces and are difficult to find and investigate; traditional anti-theft Electronic investigation methods are generally time-consuming and labor-intensive, and sometimes need to be judged based on the manual analysis of data by professionals, especially for some temporary and irregular electricity theft, it is very difficult to collect evidence

[0004] Now some units have also developed and applied some anti-stealing power systems, but due to the lack of access to data sources, the analysis ability is limited by the traditional architecture of the power consumption information collection system, and there are the following problems in the application process: First, the early warning model analysis model is single

Relying more on single abnormalities such as opening, reverse, stop, loss of pressure, and loss of flow discovered by metering online monitoring, or the method of adding index scoring and weighting, it is impossible to quantitatively evaluate the abnormal power consumption of users from a historical perspective, and it is difficult to accurately Describe the behavior characteristics of stealing electricity; second, the establishment of the model is too dependent on historical work experience

The configuration of early warning indicators is completely dependent on the electricity theft experience of historical queries, and cannot keep pace with the times. The designed parameters lack scientific and reasonable information, data analysis support, and do not have self-learning functions; third, computing power restricts the accuracy of model analysis

When building an analysis model, sampling research and production environment operation are completely different concepts. Since the algorithm or analysis technology focuses on processing a large amount of historical load, power and other data for future analysis and judgment of abnormalities, the platform cannot meet the requirements When tens of millions of users analyze and judge exceptions, the research and judgment angle is single, lack of integrity and relevance, and even cause false positives or misreports of abnormalities, which seriously affects the development of exception handling business and the practicability of functions, and cannot adapt to the Sexual Focus Screening Requirements

Images