电流型变流器SVM技术简化算法研究

SVM simplifying algorithm for current source converter

为了改善电流型变流器(CSC)的控制性能和提高控制系统的实用性,提出了一种基于反向传播神经网络(CPN)结构的空间矢量调制(SVM)技术(CPN-SVM)实现方法.该方法利用CPN的竞争学习机制,将变流器的6个非零开关矢量构成CPN的输入层,根据CPN竞争层中的胜者选择每一个采样时刻作用的开关矢量,并确定此时参考电流矢量所在的扇区,同时采用胜者的线性组合来计算SVM中所选择的开关矢量的作用时间.结果表明,与传统SVM相比,CPN-SVM避免了计算正弦函数非线性运算,大大缩短了计算时间,从而缩短最小采样周期,提高了整个系统的传输带宽,同时降低了控制系统的软硬件成本.

In order to improve the control performance and practicability of current source converter（CSC）, a novel space vector modulation （SVM） implementation algorithm based on counter propagation neural （CPN） network structure （CPN-SVM） was presented. Taking advantages of the competition and learning rules in CPN, the proposed CPN-SVM method adopted six nonzero switch vectors in CSC as input layer of the CPN. According to the winners in competitive layer, the acting switch vectors at every sample time were selected and the sector in which the reference vector lay was determined. The on-time intervals of the selected switch vectors were calculated by a linear combination of the winners. Results show that compared with the traditional SVM, the proposed CPN-SVM method avoids calculating the sinusoid function and reduces computation time, and the minimum sampling cycle is shortened and the bandwidth of the control loops is increased. The cost of hardware and software for the system is reduced.