Research On Fault Diagnosis Of High Voltage Circuit Breaker Based On Multi-information Intelligent Fusion Theory

The reliability research of high voltage circuit breakers has always been one of the hotspots of research at home and abroad.A lot of research has been done on the fault diagnosis of high voltage circuit breakers.This paper summarizes the research status at home and abroad,and proposes a multi-information fusion fault diagnosis algorithm based on this.The algorithm of weight training based on neural network is proposed to improve the algorithm.Combined with the development of artificial intelligence technology,a fault diagnosis algorithm based on deep learning self-encoder is proposed.Establish a software and hardware platform for experimental verification.This paper introduces the structure of the high voltage circuit breaker and analyze common faults firstly.Two vibration sensors and one coil current sensor are selected to collect signals during the closing process of the high voltage circuit breaker.In order to provide data support for the algorithm,it is necessary to establish a data acquisition system to collect the sensor signals in each typical state.Select hardware devices such as sensor models and build a software platform.The state of the high-voltage circuit breaker is placed on the site to simulate the state and collect and save the data.A multi-information fusion high voltage circuit breaker fault diagnosis algorithm is proposed.The sensor signal is denoised using a wavelet transform.The denoising signal is decomposed using wavelet packet decomposition,and the energy values of each frequency band after decomposition are calculated as signal feature vectors.This paper uses the acquired signal to build a fault library containing typical feature vectors.And the similarity analysis is performed to obtain the fault membership degree of each sensor for each state.The D-S evidence theory is used to fuse the membership of each sensor to obtain the fault diagnosis result.In order to solve the problem of evidence conflict in D-S evidence theory,a weighted information fusion algorithm using neural network training weights is proposed.A neural network is used as a weight calculation model to establish a connection between sensor signals and different state categories.A neural network data preprocessing algorithm is proposed and a network training rule is established.This algorithm is adopted to reduce the contradiction of evidence and improve the accuracy of information fusion.On this basis,considering the loss of fault information caused by wavelet packet transform,a multi-information intelligent fusion fault diagnosis and diagnosis system based on deep learning self-encoder is proposed.The deep-level features of the signal are extracted using a deep self-encoder to preserve the fault information as much as possible.The feature is then entered into the Softmax classifier.Algorithms such as stochastic gradient descent are employed for network training.Fault diagnosis can be performed by directly inputting the noise-reduced signal to the finally obtained system.Field experiments were performed to verify the effectiveness of the algorithm.