基于神经网络的异步电动机随机自适应动态面控制

发布时间：2018-02-16 20:17

本文关键词： 异步电动机随机非线性神经网络动态面控制输入饱和　出处：《青岛大学》2017年硕士论文　论文类型：学位论文

【摘要】：异步电动机(IM-Induction Motor)凭借低廉的制造成本,简单的结构,高度的可靠性等优点在交流调速系统和传动系统中发挥着日益重要的作用。异步电动机是一个高阶、多变量、强耦合的非线性系统,而且电机中阻尼转矩、扭转弹性转矩以及磁饱和等现象会使电机转矩、自感互感以及绕组电阻等参数发生变化,产生随机扰动,影响了电机系统的动态响应速度和控制精度。诸多学者提出了多种异步电动机驱动系统的有效控制方案,但是考虑随机扰动的异步电动机传动系统的控制策略研究还相对较少。因此,研究适用于异步电动机随机系统的控制方法,提高异步电动机调速系统的动静态性能具有重要的理论意义和实际应用价值。本文结合动态面和反步法研究了异步电动机随机系统的神经网络速度调节控制问题。利用径向基函数(RBF)神经网络来逼近系统中未知的非线性函数,结合动态面技术和反步控制构建非线性控制器,有效地消除了随机扰动的影响,实现了对异步电动机调速系统的高品质控制。论文的主要研究成果可以概括如下:1.研究了基于神经网络的随机非线性系统自适应动态面控制问题。使用RBF神经网络逼近系统的非线性项,引入动态面通过其一阶低通滤波作用避免了导致控制器结构复杂的“计算爆炸”问题,克服随机扰动的影响,根据反步原理构造整个系统的自适应控制器。最后由Lyapunov方法分析了该方法的稳定性。2.基于动态面技术和神经网络原理,采用自适应反步法设计了异步电动机随机系统的速度调节控制策略。利用神经网络对系统非线性项做逼近处理,动态面技术的运用有效避免了传统反步设计中普遍存在的“计算爆炸”问题;整个系统的真实控制律在反步控制的最后一步给出,稳定性分析表明所构造的控制器能够克服随机扰动的不利影响,使系统内所有的信号都保持有界。仿真实验的结果证明该控制器调速效果优良,且具有较强的鲁棒性。3.将输入饱和限制引入异步电动机的随机系统模型,同样使用RBF神经网络处理系统非线性项,动态面的运用大大减少了控制器构造过程中的计算量,弥补了传统反步设计的不足。最终构建出能使异步电动机实现良好调速效果的控制器,克服了随机扰动的不利影响。该控制器只有一个自适应参数需要调节,且考虑了输入饱和的影响,因此更加具有实际使用价值。
[Abstract]:Induction Motor (IM-Induction Motor) is playing an increasingly important role in AC speed regulation system and transmission system by virtue of its low manufacturing cost, simple structure and high reliability. Strong coupling nonlinear system, and the damping torque, torsional elastic torque and magnetic saturation in the motor will make the motor torque, self-inductance mutual inductance, winding resistance and other parameters change, resulting in random disturbance. The dynamic response speed and control accuracy of motor system are affected. Many scholars have put forward many effective control schemes for asynchronous motor drive system. However, there is relatively little research on the control strategy of the asynchronous motor drive system considering the random disturbance. Therefore, the control method suitable for the asynchronous motor drive system is studied. It has important theoretical significance and practical application value to improve the static and static performance of asynchronous motor speed regulating system. In this paper, the neural network speed regulation control problem of asynchronous motor stochastic system is studied by combining dynamic plane and backstepping method. The radial basis function (RBF) neural network is used to approximate the unknown nonlinear function in the system. Combined with dynamic surface technique and backstepping control, a nonlinear controller is constructed, which effectively eliminates the influence of random disturbance. The main research results of this paper can be summarized as follows: 1. The adaptive dynamic surface control problem of stochastic nonlinear system based on neural network is studied. RBF neural network is used. The nonlinear term of the system of approximating a network, The introduction of dynamic surface through its first-order low-pass filter avoids the problem of "computational explosion", which leads to the complex structure of the controller, and overcomes the influence of random disturbance. The adaptive controller of the whole system is constructed according to the backstepping principle. Finally, the stability of the method is analyzed by the Lyapunov method. 2. Based on the dynamic surface technique and the neural network principle, The adaptive backstepping method is used to design the speed control strategy for the asynchronous motor stochastic system. The nonlinear terms of the system are approximated by the neural network. The application of dynamic surface technology effectively avoids the problem of "computational explosion" in traditional backstepping design, and the real control law of the whole system is given in the last step of backstepping control. The stability analysis shows that the proposed controller can overcome the adverse effects of random disturbances and keep all the signals in the system bounded. The simulation results show that the controller has a good speed regulation effect. And it has strong robustness. 3. The input saturation limit is introduced into the stochastic system model of asynchronous motor. RBF neural network is also used to deal with the nonlinear terms of the system. The application of dynamic surface greatly reduces the calculation in the process of controller construction. It makes up for the shortcoming of the traditional backstepping design. Finally, a controller which can make the asynchronous motor achieve good speed regulation effect is constructed, which overcomes the adverse effect of random disturbance. The controller has only one adaptive parameter to be adjusted. The effect of input saturation is considered, so it has more practical value.
【学位授予单位】：青岛大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP273;TM343

【参考文献】