Three level inverter fault diagnosis using EMD and support vector machine approach

A fault diagnosis strategy for neutral point clamped three level inverter (NPC) is proposed. Due to the non-stationary characteristics of fault signals, empirical mode decomposition and support vector machine (EMD-SVM) are adopted in the strategy. The load phase voltages are selected as fault detection signals, after preprocessed using the empirical mode decomposition (EMD), fault signals features are extracted by principal component analysis (PCA) with reduced samples dimensions. Finally, an SVM model is used to classify these faulty samples. Simulation results prove the feasibility and good classification performance of the proposed EMD-SVM strategy.