One statistics-based fault classification technique for cascaded inverter

A fault classification and location method for a 5-level Cascaded H-Bridges Inverter based on BP Neural Network is proposed in this paper. Meanwhile, Principal Component Analysis (PCA) is used in the Neural Network,which we call PCA-NN, to simplify the training data and save the training time. On the other hand, a BP Neural Network without PCA, which is called NN in the following sector, is also proposed. Simulation is done through MATLAB to certify the feasibility of the proposed networks. At the same time, a contrast is made to compare the performances of the two Neural Networks and conclusions are drawn. From the contrast it can be seen that PCA-BP is 5 percent more accurate than NN and the PCA-NN considerably cut down the dimension of training data from 30 to 7 which is favorable for saving training time and improving mapping performance.