基于BP神经网络的碳化硅MOSFET栅极老化监测方法研究

Research on Gate-oxide Degradation Monitoring Method of SiC MOSFET Based on BP Neural Network

针对碳化硅MOSFET运行工况复杂、易造成器件栅极老化、影响电力系统可靠性的问题,提出一种基于BP神经网络的碳化硅MOSFET栅极老化监测方法。以碳化硅MOSFET的阈值电压和体二极管通态压降作为栅极老化的敏感表征参数,设计、搭建测试实验平台,获取变测量条件下的电参数值,结合BP神经网络提取健康器件与老化器件样本数据间的特征差异,充分挖掘器件的可靠性信息。实验结果表明:该方法可对碳化硅MOSFET的栅极老化状态进行较为准确的检测和评估。

Aiming at the problem of gate-oxide degradation of SiC MOSFET caused by complex operating conditions,which could affect the reliability of the power electronic systems,a BP neural network-based gate-oxide degradation monitoring method of SiC MOSFET was proposed.The threshold voltage of SiC MOSFET as well as the on-state voltage of the body diode were chosen as the sensitive parameters of gate-oxide degradation,and a test experimental platform was designed and built to obtain the two parameters under variable measurement conditions.The BP neural network was used to extract the differences between the healthy devices and the degraded devices,and fully explore reliability information of the devices.The results show that this method can accurately test and evaluate the gate-oxide degradation state of SiC MOSFET.