A Power Transformer Fault Diagnosis Method Based on Improved Variational Quantum Shadow Learning

An improved variational quantum shadow learning (VQSL) method is presented for the fault diagnosis of transformers. Localized variational quantum circuits for shadow feature extraction and a fully connected neural network are designed. Firstly, the dissolved gas analysis data is mapped into a high-dimensional Hilbert space, then, shadow features between adjacent and non-adjacent quantum bits are extracted simultaneously by multiple localized quantum circuits through parameter sharing, which reduces the number of training parameters and improves the model accuracy. The key factors that influence the performance of the VQSL method are discussed, which include quantum coding methods, the width and the depth of localized quantum circuits, the size of the global quantum bits, and the noise levels. Experimental results indicate that the improved VQSL method achieves a diagnostic accuracy of 95.4%, surpassing the commonly used methods such as QNN, GNN, CNN, SAE, SVM, and ratio methods. Even in a noise interference situation, the proposed method consistently maintains a high fault diagnosis accuracy.