Artificial Neural Network Based Trilogic SVM Control in Current Source Rectifier

Various modulation methods for the Current Source Rectifier (CSR) controlling scheme have been investigated in recent years. The traditional modulation methods have the disadvantages such as the great computing cost, sensitivity to load and system parameter variation. In this study, an Artificial neural network (ANN) based algorithm is adopted to tackle the problem. This algorithm features parallel computation and self-tuning. The Random weight change (RWC) algorithm is employed for on-line parameter tuning to achieve better performance. The principle of the trilogic Space vector modulation (SVM) for CSR is introduced as the theoretical foundation. The proposed method is introduced in two parts, the constructing of the neural network and the designing of an on-line parameter tuning algorithm. The simulation results based on SABER software show that the new algorithm has a good performance, especially under a non-rated system load.