基于神经网络技术的电机直接转矩控制系统研究

Study on motor direct torque control system based on neural network technology

传统的电动机直接转矩控制系统存在一些问题,比如转矩脉动较大、磁链观测不精确、开关频率不固定等,影响了控制系统整体性能的进一步提升。为此,结合神经网络技术,设计了近似圆形磁链的直接转矩控制系统优化方案,探讨采用BP神经网络实现电压矢量选择的功能。该方案采用滞环比较器对磁链进行调节并实现转矩调节,建立了空间电压矢量选择的最优方式,设计了神经网络空间电压矢量选择器结构,提出采用神经网络理论取代传统开关表的方法。系统仿真结果表明,优化后的系统能够有效降低转矩脉动,定子磁链轨迹的脉动成分也明显下降,在低速运行状态下同样取得了较好的效果,表明该系统达到了调速性能改善和整体性能提升的目的。

Since traditional motor direct torque control system has some problems, such as large torque ripple, inaccu- rate flux linkage observation, unfixed switching frequency, which influence on the integral performance improvement of control system further. Therefore, in combination with neural network technology, the direct torque control system optimization scheme which approximates circular flux linkage was designed. Voltage vector selection function realized by BP neural network is dis- cussed. The system adopts hysteresis comparator to adjust the flux linkage and achieve torque control. The optimal manner for space voltage vector selection was established, then the structure of space voltage vector selector was designed. The method of using neural network theory instead of traditional switching table is proposed. System simulation results show that the optimized system can reduce torque ripple effectively, and the ripple component in stator flux trajectory is also decreased significantly. Good effect is acquired in low-speed operating condition. The system achieves the improvement of speed regulation performance and the promotion of the integral performance.