Principal component analysis (PCA) based sparrow search algorithm (SSA) for optimal learning vector quantized (LVQ) neural network for mechanical fault diagnosis of high voltage circuit breakers

Abstract

An essential piece of equipment for the protection and management of the electricity grid is a high voltage(HV) circuit breaker. The reliability and stability of the electricity grid are directly affected by the condition of operation of a high voltage circuit breaker. This study proposes a PCA-SSA-LVQ-based high-voltage circuit breaker diagnosis method to identify mechanical faults of high-voltage circuit breakers more quickly and effectively. Laboratory simulation of vibration signals for different fault states of high voltage circuit breakers. Finite intrinsic mode functions(IMF) are obtained by empirical modal decomposition(EMD) of the simulated vibration signal. The envelope spectrum of the IMF is extracted using the Hilbert envelope transform, and the concept of information entropy is used to determine the entropy value of the envelope as the characteristic quantity of the signal. Using principal component analysis, one can decrease the eigenvectors’ dimension. The high voltage circuit breaker LVQ neural network fault diagnosis model is made, and the initial weight of the LVQ neural network is optimized using the SSA optimization algorithm. The accuracy was improved in the end. The optimized and non-optimized LVQ neural networks were compared. The results show that the PCA-SSA-LVQ neural network has a higher diagnostic rate and faster training convergence. The main contribution of this paper is to reduce the redundancy of feature vectors using the PCA algorithm, optimize LVQ neural network by the SSA algorithm, and creatively use these three algorithms together. The training speed and the accuracy of state recognition are improved.