Switchgear reliability evaluation method

Abstract

The invention provides a switchgear reliability evaluation method which comprises the following steps: (S1) pretreating and initializing the basic data and operation data generated by a switchgear respectively, and establishing corresponding vector comparison tables; (S2) establishing a tree structure of the basic data, and modeling by use of a recurrent neural network; (S3) establishing a time sequence structure of the operation data, and modeling by use of a circular neural network; (S4) establishing a loss function according to the models established in the step (S2) and step (S3), and performing model training; and (S5) in combination with the vector comparison tables, performing model-based reasoning to obtain final risk assessment. According to the invention, by combining the recurrent neural network and the circular neural network and introducing an attribute vector and an event vector, the problem of low-dimensional coding of nonnumeric data is solved; and in combination with the basic data and the operation data, the fusion of multi-source information can be realized so as to provide accurate quantitative evaluation and dynamic evaluation of the switchgear reliability.

Description

technical field [0001] The invention relates to the technical field of power systems, in particular to a method for evaluating the reliability of switchgear. Background technique [0002] Switchgear is the key control and protection equipment in the power system. Once it fails, it will cause a power outage in the community under its jurisdiction and even the entire area. In addition to the need to invest a lot of manpower and material resources during power outage maintenance, it also requires a long maintenance time, which will inevitably cause huge economic losses and social impact. The estimation of the probability of failure, that is, the reliability evaluation is of great significance to reasonably arrange the maintenance time and reduce the risk of failure. [0003] At present, there are several methods for evaluating the reliability of switchgear: 1. Use expert experience to judge and combine the design service life of the equipment to estimate the remaining life. So...

AI Technical Summary

Problems solved by technology

A large amount of valuable information, such as equipment model, environmental factors, family defect history, operation history, load and other information are not well utilized

Secondly, the statistical analysis needs to obtain a more accurate evaluation under the condition of a large sample. First, the equipment models are different and the operating conditions are different, so it is difficult to grasp the balance between the statistical granularity and accuracy; second, the statistical analysis under a large sample The value has limited guiding significance for the operation and maintenance of a single device; in the third method, the object of online monitoring is often a specific value, which inevitably requires more online monitoring values ​​for the overall reliability of the device, which is of great importance to cost and safety. big test

In addition, in the actual operation of the network, the coverage of online monitoring equipment has not reached an ideal level, and the identification of some online monitoring objects still needs further research; in the fourth method, due to the shallow neural network structure Its learning ability is limited, and there are many factors affecting the reliability of switchgear. Neural networks and other methods may not be able to obtain ideal results in the face of high input dimensions, and if the number of inputs is low, it cannot reflect the complete status of the equipment.

In addition, the above methods also face a common problem that they face the difficulty of complex coding. A large number of variables in the power system exist in non-numerical forms, such as substation names, switchgear models, manufacturers, etc., can only use very high coding dimensions to distinguish each other, and can only use simple statistical methods to compare different performances, and cannot carry out deeper knowledge mining

Images