摘要
在直接转矩控制系统中,感应电机定子电阻在低速时受温度影响而变化,影响控制系统的低速性能。为此,设计了基于反向传播(BP)神经网络的定子观测器,采用有导师学习(或称监督学习)的神经网络学习方法,通过试验,采集到定子电阻随其电压频率及定子绕组温度变化规律,利用MATLAB人工神经元网络工具箱就能快速实现BP计算,将其结果输入定子电阻观测器中,进行实时检测,计算出准确的定子磁链幅值,达到减小系统低速运行时转矩的脉动,提高了控制性能。
In direct due to the effect torque control system, the stator resistance of induction-motor in a low speed, is variable of temperature. Consequently, the low speed performance of the control system is influenced. In order to solve this problem, a stator resistance intelligent observer is designed based on BP neutral network, which adopts suppervised parameter learning in neutral network. Through experiment, the rule by which the stator resistance is variable with the voltage frequency and the stator loop temperature can be got. Using the neural network tool of the Matlab, the BP arithmetic can be realized rapidly. It can reduce the torque fluctuation in low speed and improve the control capability through putting the trained result into the observer, surveying real time data and computing the accurate amplitude value of the stator flux linkage
出处
《电力系统及其自动化学报》
CSCD
北大核心
2008年第3期25-28,共4页
Proceedings of the CSU-EPSA
基金
教育部科学研究资助项目(204032)
关键词
直接转矩控制
BP神经网络
观测器
仿真
direct torque control
BP neutral network
observer
simulation
