Random-forest-based circuit breaker mechanical fault diagnosis method and system

Abstract

The invention discloses a random-forest-based circuit breaker mechanical fault diagnosis method and system. The method comprises: establishing a circuit breaker vibration sample database; respectivelyextracting time domain feature vectors of each vibration sample data in the circuit breaker vibration sample database, and combining the time domain feature vectors to obtain a feature vector; forming an original training sample set; collecting vibration signal data of a target circuit breaker, and preprocessing the data; extracting time domain feature vectors of vibration signal data of the target circuit breaker; combining the obtained time domain feature vectors into a feature vector F; and according to the obtained original training sample set and the feature vector F, using a random forest algorithm to perform fault diagnosis on the target circuit breaker. The method has the beneficial effects that the introduction of two kinds of randomness makes the random forest have good anti-noise ability, is especially suitable for fault diagnosis of circuit breakers, and can improve the accuracy of fault diagnosis.

Description

technical field [0001] The invention relates to the detection field of high-voltage electrical equipment in power systems, in particular to a method and system for diagnosing mechanical faults of circuit breakers based on random forests. Background technique [0002] As an important protection and control electrical appliance in the power system, medium and high voltage circuit breakers play a key role in ensuring the safe and stable operation of the power grid, and their operation and maintenance are also an important part of the daily work of the power sector. [0003] According to the statistics of the national power system distribution voltage level switch accidents from 1990 to 1999, mechanical failures accounted for 39.30% of the total failure types. In recent years, various data mining algorithms have been widely used in the mechanical fault diagnosis of medium and high voltage circuit breakers, and have achieved good results, such as the diagnosis system based on art...

AI Technical Summary

Problems solved by technology

However, the artificial neural network has shortcomings such as difficult parameter optimization and slow convergence speed. Although the support vector machine overcomes the slow convergence speed and over-fitting problems of the artificial neural network, it also has insufficient ability to deal with large sample data and solve multiple problems. Difficulties such as low accuracy of classification problems

[0004] To sum up, in the existing data mining algorithms, there is still a lack of effective solutions for how to efficiently improve the accuracy of mechanical fault diagnosis.

Images