【作者】 修杰；

【导师】 夏长亮；

【作者基本信息】 天津大学 ， 电机与电器， 2007， 博士

【摘要】 开关磁阻电机驱动系统(SRD)是一种新型交流驱动系统,以结构简单、坚固耐用、成本低廉、控制参数多、控制方式灵活、可得到各种所需的机械特性,而备受瞩目,应用日益广泛,并且SRD在宽广的调速范围内均具有较高的效率,这一点是其它调速系统所不可比拟的。由于开关磁阻电机(SRM)磁路高度饱和、结构特殊,存在严重的局部饱和、漏磁、边缘效应等现象,传统的等效磁路分析方法很难准确分析SRM的磁场分布。电磁场有限元(FEM)数值分析是分析SRM磁场分布的有效工具。本文采用电磁场有限元分析了SRM的磁场分布,得到磁密分布、等矢量磁位线分布并计算了磁链ψ(θ,i)在不同转子位置、不同电流下的磁化曲线族,为SRM的精确分析、计算打下了基础,并为磁路的合理设计提供了参考。SRM特殊的双凸极结构及磁路的高度饱和,使得对其分析、计算十分困难。SRM的磁链为转子位置及绕组电流的非线性函数,建立这一非线性映射关系成为精确分析、计算SRM特性的基础。神经网络具有任意非线性函数逼近能力、较强的学习能力、自适应能力。Takagi-Sugeno(T-S)型模糊逻辑可充分利用专家知识及语言信息,其后件为输入变量的线性组合,具有计算简洁、运算速度快、精度高的特点。本文采用综合了两者优点的pi-sigma模糊神经网络建立了SRM的磁特性非线性模型,具有鲁棒性强、容错能力强、精度高等特点,仿真结果证明了这一点。准确而实时的转子位置信息是SRM运行的必要信息,传统的位置信息是由机械式位置传感器提供的。位置传感器是传统SRM的一个标志性部件。机械式位置传感器的存在增大了SRM的体积、提高了成本、增加了制造的复杂程度、降低了系统的可靠性,因此无位置传感器的SRM控制方式成为研究热点。本文提出采用模糊自适应神经网络(ANFIS)来映射转子位置与绕组磁链和绕组电流之间的非线性关系,在检测到绕组磁链和绕组电流后,经ANFIS运算得到转子位置角,这一方法的优点是具有较强的容错性、对噪声信号有较强的抑制能力、精度高、鲁棒性强。实验结果证明本方案是一种较好的SRM无位置传感器位置检测方法。开关磁阻电机因磁路的饱和导致参数的非线性,又因在不同控制方式下是变结构的。这使得开关磁阻电机的控制非常困难。经典的线性控制方法如PI、PID等方法用于开关磁阻电机的控制很难取得较好的控制效果。其它的控制方法如滑模变结构控制、状态空间控制方法等可取得较好的控制效果但大都比较复杂,实现起来比较困难。作为智能控制分支之一的神经网络控制,因具有任意非线性逼近能力、自学习能力、自适应能力,故对非线性、不确定、不确知、变结构、时变的被控对象可取得较好的控制效果且不需知道被控对象的数学模型,这对于很难精确建模的开关磁阻电机来说尤其适用。本文将神经网络与PID控制相结合充分发挥神经网络的自适应、非线性映射能力和学习能力,提出了一种自适应能力很强的参数可调的神经网络PID控制策略。同时采用RBF神经网络建立系统的非线性预测模型,进行参数预测,提高了系统的动态响应特性,既具有PID控制精度高、实现容易的特点,又具有神经网络的自适应特性,对具有很强非线性特性的SRD取得了较好的控制效果,动态响应快、超调小、鲁棒性强。更多还原

【Abstract】 Switched Reluctance Drive (SRD) system is a novel AC driving system. It has a characteristic of simple structure, robustness, low cost, reliability and control flexibility. It has many control method and can get varity machinery characteristic. Also, it has a high efficiency in a wide speed range. This is super to other adjustable speed driver system.For its high saturation of magnetic circuit and special double salient structure and high saturation of local magnetic circuit, the traditional circuit analysis method is no longer suitable to be used to calculate the distribution of flux of SRM. Magnezation finite element method (FEM) is a powerful tool to calculate the distribution of flux. In this paper, FEM is used to analysis the distribution of flux of switched reluctance motor (SRM). The distribution of magnetic dense and the equipotential lines of the SRM used in this paper are obtained. Also, the flux linkageψ(θ,i) is calculated at different current and different rotor position. This is the base to calculate the performance of SRM accurately. The analysis also provides a reference to design the magnetic circuit.The highly saturation of magnetic circuit and the doubly salient structure of SRM lead to flux linkage is in nonlinear function of both rotor position and phase current. Building up this nonlinear mapping is the base to calculate the property of SRM accurately. Artificial neural networks (ANN) under certain condition can approximate any nonlinear function with arbitrary precision which also has a strong learning ability and adaptive ability. While Takagi-Sugeno (T-S) type fuzzy logic which antecedents are fuzzy sets and consequents are linear combination of the input variables, the output of which are crisp values. So the inference process of it can be simplified. And it can take advantage of the numerical information and language information. So, in this paper, one form of T-S type fuzzy logic– pi-sigma neural networks is adopted to develop the nonlinear model of SRM. By taking advantage of the benefit of neural network and fuzzy logic, a high precision model of SRM with a characteristic of robustness, error tollerance, and high precision is developed. The simulation results proved this.The accurate and real-time rotor position information is very important for high performance operating of SRM. Traditionally, the rotor position information is provided by a mechanical rotor position sensor. But this increases cost, size and manufacture complexity of SRM and reduces the reliability of the system. So the sensorless control method of SRM brcome a hot research region. Adaptive network based fuzzy inference system (ANFIS) is used in this paper to map the nonlinear function of rotor position with respect to flux linkage and phase current. After the flux and current are measured, the rotor position is computed by the ANFIS. The advantage of this method is that it has strong error tolerance ability, high noises restrain ability, high accuracy and robust. The experimental results proved the effectiveness of the proposed method.The high saturation of magnetic circuit leads to the nonlinear of parameter of SRM and under different control method it is variable structure. This makes it hard to get a good performance by using the conventional PI, PID controller to the speed control of SRM. Other control method such as sliding model control, state-space control method can get a good control performance. But they are too complexity. ANN control - one of intelligent control, under certain condition, can approximate any nonlinear function with arbitrary precision. It also has a strong ability of self-learning and adaptive. So it can be used to control nonlinear, uncertain, unknown, variable structure, time varing and methmatical model unknown plant. By combining it with the conventional PID controller, an adaptive PID controller is developed in this paper. Meanwhile nonlinear prediction model based on a radial basis function (RBF) ANN is build up to predict the parameter of the system. This improves the dynamic response of the system. This control method has the advantage of PID controller’s high precision and implement easily. It also has the advantage of ANN’s characteristic of adaptive. Appling it to the speed control of SRM, a good control performance is got. The system responds quickly with little overshot and is robust.更多还原