基于神经元的异步电机DTC系统转速辨识

【作者】 高宏洋；

【导师】 喻寿益；

【作者基本信息】 中南大学 ， 电力系统及其自动化， 2008， 硕士

【摘要】 在高性能的直接转矩控制系统中,速度闭环是必不可少的。传统的DTC系统多采用光电脉冲编码器等速度传感器反馈转速信号。但是,安装速度传感器会降低系统可靠性,增大电机体积,增加系统成本。因此,基于无速度传感器的直接转矩控制技术成为目前交流传动的研究热点。论文针对以往转速辨识方法的不足,提出了以定子磁场定向的基于神经元的异步电机直接转矩控制系统的转速辨识方法。该方法以直接转矩控制理论为基础,在静止两相α-β坐标系下建立异步电机数学模型,推导出定子电流状态方程,并转化为带有5个输入节点、1个输出节点的神经元,建立了以定子电流为驱动的自适应辨识可调模型。采用Window-Hoff学习规则,推导出速度辨识公式,把辨识出的速度估计值反馈到可调模型中,在线修改可调模型参数,使其输出的定子电流估计值逼近于实时检测的定子电流值,产生自适应性,提高了定子电流和电机转速的辨识精度。使用37kW异步牵引电机参数在matlab/simulink软件平台上进行仿真,并与传统的基于神经元网络的转辨识方法作比较。仿真结果表明,辨识转速能够准确地跟踪实际转速,动态响应性能好,消除了传统方法中转速下陷的问题,且定子磁链更加接近于圆形。尤其在参数发生突变时,系统具有很好的鲁棒性,并且实现简单。论文还以TI公司的DSP控制器TMS320F2812为核心,对无速度传感器直接转矩控制系统进行了全数字化设计,实现了采用改进算法的无速度传感器直接转矩控制系统,并在硬件和软件两方面进行了介绍。更多还原

【Abstract】 The closed-loop of speed is essential in the high-performance direct torque control system.These sensors such as the photoelectric pulse encoder had been used for the feedback of speed signal in the traditional DCT system. However, installing the speed sensor brought about series of problems such as lower reliability、bigger volume and higher cost. The direct torque control technology based on speed sensorless had become the current researching hotspot of AC drive.A new kind of speed identification method which was based on a stator magnetic field oriented and neurons theory in the DTC system was presented in this dissertation against the deficiencies of the prevenient speed identification method. Asynchronous motor mathematical model was established in statorα-βcoordinate system according to the new method based on DTC theory, and stator current equation of state could be gained according to this model and transformed into neurons with five input nodes—one output node, an adaptive identification modules driverred by stator current was derivated. According to speed identification formula, speed could be estimated fleetly and transferred back to adjust model whose parameters could be modified, and the adjustment made them approximate to the real value. The simulation results compared with the conventional speed identification method based on neurons in the matlab/sinulink using parameters of 37kW asynchronous motors showed the identified speed could accurately track the actual speed, perform good dynamic response、eliminate the speed settlement of the traditional methods and the stator flux is rounder. In particular, when the parameters of system changed suddenly, the system was of great robustness and easy to application.A digital system for speed sensorless DTC system had been designed based on TMS320F2812 digital processor which is produced by the TI company. The improved speed-identification algorithm was implemented in this system against the traditional DTC system. There was a introduction to the hardware and software of the system and some programs.更多还原