基于线性神经网络的单相SPWM逆变器畸变抑制

Distortion Suppression Based on Linear Neural Network for Single-phase SPWM Inverter

正弦波脉宽调制(SPWM)逆变器受功率开关器件固有的死区时间、导通压降以及负载变化等因素影响导致输出发生畸变,从而引发设备运行稳定性降低、损耗变大等严重问题。从各种类型畸变的产生机理着手,针对死区效应,给出了实现死区时间最小化、减轻输出过零畸变的算法。针对模型参数、负载变化对补偿效果的影响,提出了基于自适应线性神经网络(Adaline)的畸变检测、补偿算法。最后,在采用STM32F103ZE6控制器的单相SPWM逆变器上进行了实验验证,结果表明输出畸变为0.9%,较补偿前降低约85%。

By power switching devices dead time,voltage drop and load,the output of SPWM inverter has distorted,a series of problems like reducing stability of equipment,inceasing loss and so on are presented.From the reasons of different types of distortion,a algorithm to achieve minimum dead time and to reduce the distortion of zero crossing is given.Aiming at the change of model parameters and load,a algorithm of distortion detection and compensation based on Adaline is proposed.Finally,using STM32F103ZE6,single-phase SPWM inverter results show that the distortion of output is only 0.9%,lower than about 85% before compensation.