【作者】 王迎发；

【导师】 夏长亮；

【作者基本信息】 天津大学 ， 电机与电器， 2007， 硕士

【摘要】 无刷直流电机(BLDCM)体积小、重量轻、效率高,又克服了有刷直流电机机械换向带来的一系列缺点,因此在各个领域得到广泛应用。针对本身具有时变性、非线性、强耦合等特征的无刷直流电机,高性能的调速控制方法成为一个重要的研究方向,本文研究了智能控制算法在无刷直流电机控制中的应用。PID控制是工业控制中最常用的方法,但用其对具有复杂非线性特性的对象或过程进行控制难以达到满意的效果。神经网络控制具有非线性映射能力、学习能力和自适应能力,是一种不依赖模型的控制方式,但是,神经网络的结构和初始值的选取影响控制器的性能,采用反复试验初始值的方法很难得到最优参数的控制器,这影响了该控制器的广泛应用。由Holland首先提出的遗传算法是模拟生物在自然环境中的遗传和进化过程而形成的一种随机搜索的全局优化算法,该方法可以全局搜索得到最优参数。基于遗传算法的控制器的参数优化成为控制器设计的一种有效方法。针对径向基函数(RBF)神经网络的结构确定和隐层节点参数调节的不足,本文提出了一种基于遗传算法训练RBF神经网络控制器的无刷直流电机自适应控制方法。该方法利用遗传算法离线训练RBF神经网络的结构和隐层节点参数,构成速度控制器,在电机运行中通过控制器的在线修正,自适应地调节神经网络参数,同时电流环迅速跟踪给定电流的变化,达到使系统适应环境变化的目的。MATLAB仿真结果表明,该方法响应速度快、控制精度高,同时具有适应性好、鲁棒性强等优点。最后设计了以TMS320F2812为核心的电机调速系统硬件电路,对无刷直流电机调速系统进行了研究。更多还原

【Abstract】 The brushless DC motors (BLDCM) have small sizes, low weight, and large power to volume ratio, and receive much attention in every field. Compared with brush DC motors, BLDCM overcome a series drawbacks brought by mechanical converted direction. The BLDCM is a multi-variable and non-linear system, so the research about the high performance of speed regulation becomes an important direction. This paper researches the application of intelligence algorithm in the control of BLDCM system.Classical PID control is most common method in industry, but this method can’t gain satisfying effect to the complex and non-linear object or process. Neural network control has the ability of expressing arbitrary nonlinear mapping, learning and self-adaptive. The selection of topology structure and initial value influence the capability of controller, and the method of trial and error has much difficult in getting optimization parameter controller. Genetic algorithm first put forwarded by Holland is a global optimization algorithm of random search that simulate the process of inheritance and evolution formed in natural condition. Parameter optimization based on genetic algorithm becomes an effective method in designing controller.To solve the deficiency of Radial Basis Function (RBF) neural network such as decision of structure and adjustment of parameters in hidden-unit, this paper presents an adaptive speed control approach based on genetic algorithm. In this approach, the RBF neural network whose structure and parameters of hidden-unit have been trained by genetic algorithm off-line constitutes a speed loop controller. The controller tunes parameters of neural network adaptively via the self-modifiability of network on-line. At the same time, the current loop controller traces the change of given current rapidly, so that the system can adapt to the variational environment. The results of MATLAB simulation prove that the approach has lots of good performances in response speed, control accuracy, adaptability and robust. Finally, a hard circuit based on digital signal processor (DSP) TMS320F2812 is designed to further research speed regulation system of BLDCM.更多还原