Fault Diagnosis of Power Transformer Based on KPCA-RF

To accurately diagnose faults in power transformers, this paper proposes a method for power transformer fault diagnosis that utilizes Kernel Principal Component Analysis (KPCA) and Random Forest (RF). Initially, the method involves expanding the data dimensions by calculating rations among five common Dissolved Gas Analyses (DGA). Following, critical feature data is extracted using KPCA to reduce data redundancy. Subsequently, select 10 principal components to serve as input for the RF model to predict transformer faults. Simulation demonstrates a certain improvement in accuracy compared to traditional algorithms, thus holding significant practical implications in engineering applications.