Circuit breaker energy storage process state identification method for constructing CNN characteristic matrix by acoustic vibration signal

Abstract

The invention discloses a circuit breaker energy storage process state identification method for constructing a convolutional neural network (CNN) characteristic matrix by an acoustic vibration signal. The method comprises the following steps: firstly proposing a time scale alignment method based on kurtosis and envelope similarity to guarantee the synchronism of the acoustic vibration signal; secondly, detecting a vibration signal starting point by adopting a Lyapunov exponential wavelet modulus maximum value (Lwavelets), and constructing an acoustic vibration signal two-dimensional feature matrix by utilizing a Pearson correlation coefficient after overlapped data expansion is carried out on the data; finally, training the feature matrix by using the CNN, optimizing the CNN structure byusing a support vector machine (SVM) instead of a SoftMax classifier, and searching SVM optimal parameters by using grey wolf optimization (GWO). The optimized CNN model is not sensitive to the situation that the data change of the circuit breaker energy storage process is large; as a new circuit breaker energy storage process state recognition method, the method greatly improves the accuracy andgeneralization of state recognition.

Description

technical field [0001] The invention relates to the technical field of fault diagnosis of electrical equipment, in particular to a method for constructing a feature matrix required by a convolutional neural network (CNN) by combining sound and vibration signals, and then identifying the state of the energy storage process of a circuit breaker. Background technique [0002] As an important control and protection device in the power system, the reliability of the circuit breaker directly affects the safety and stability of the power system. Therefore, the reliability of the circuit breaker is very important to the protection and control of the power grid. [0003] At present, the research on fault diagnosis of circuit breakers is mostly focused on the opening and closing process: using the control coil current, insulation rod displacement, and vibration signals to identify mechanical faults. The research focus is on the problems that occur in the equipment itself during the ci...

AI Technical Summary

Problems solved by technology

However, the traditional sound-vibration signal combination method does not consider the difference between the two, and combines the sound-vibration signals mechanically. Although deep learning algorithms such as convolutional neural networks are introduced, due to the lack of feature information and the general structure of CNN, the pertinence is poor. , resulting in insufficient diagnostic accuracy and poor generalization

Images