【作者】 高洁；

【导师】 孙鹤旭；

【作者基本信息】 河北工业大学 ， 电气工程， 2012， 博士

【摘要】 开关磁阻电机（SR电机）的转矩脉动及其振动、噪声是SRD系统的显著缺点，如何抑制转矩脉动历来都是SRD研究的难点与热点问题，目前关于控制优化研究的较多，而关于电机本体优化的研究相对较少，且已有的控制大多未考虑互感对转矩的影响。针对上述问题，本文在综合分析SR电机转矩脉动机理的基础上，结合粒子群算法快速收敛、全局寻优的优点，主要通过对电机结构的优化设计、功率变换器拓扑结构的改进、瞬时转矩补偿控制策略的应用三方面共同实现转矩脉动的全局优化。首先通过推导SR电机输出方程进行了系统完整的结构设计，尤其是针对应用最广泛的四相8/6极结构，利用稳态有限元法分析了计及互感耦合、饱和效应，且考虑因单、双相励磁模式以及长、短磁路连接方式差异对磁链特性、电感特性、矩角特性的影响。实测了一台SR样机的自感及互感特性，验证了有限元仿真的准确性。建立了SR电机场-路-运动耦合分析模型，分别利用时步有限元法和实验对SR电机从起动到稳态运行特性进行了详细研究。其次从电机本体结构设计的角度考虑，研究如何从源头对其性能进行优化，通过有限元法分析电机磁极形状对目标函数的影响，得出与之相关的优化变量和约束条件。利用广义回归神经网络的训练获取了目标函数的数学模型，并且通过自适应粒子群算法求解非线性、多变量、多约束条件的SR电机优化设计问题，计算程序可推广至任意相数、任意结构的电机，为SR电机的优化设计提供了一种通用算法。同时从如何优选电机控制参数及优化功率变换器拓扑结构的角度出发，在推导计及互感的SR电机两相导通基本方程并采用Simulink非线性建模的基础上，深入研究开关角、斩波频率对相电流及转矩脉动的影响，为合理选择控制参数提供了有力依据；基于电容补偿原理，提出一种两相励磁功率变换器拓扑结构的改进方案，减小转矩脉动的同时提高了系统功率因数及效率。最后基于粒子群-广义回归神经网络算法求解转矩逆模型，实现了SR电机瞬时转矩控制，且针对两相励磁模式，考虑到偶数相电机长磁路连接互感较大的缺陷，特别加入互感转矩补偿模块，对计及互感转矩的SR电机低转矩脉动调速系统进行了仿真研究，有效减小了因互感产生的负转矩，转矩脉动显著减小。搭建了SRD系统实验平台，分别分析了常规PWM控制及瞬时转矩控制对相电流、输出转矩及其脉动的影响，并且依次进行了稳态运行、动态运行的实验研究，结果表明瞬时转矩控制策略使得相电流更平稳，鲁棒性更强，抗干扰能力更好，实现了SR电机转矩脉动的全局优化控制。更多还原

【Abstract】 The instantaneous torque and vibration, noise caused by ripple of Switched Reluctance(SR) motor are not only more prominent problem, but also the difficulties in SRD research. Atpresent how to suppress the torque ripple has become a hot spot, but the study focuses more onthe control optimization and the motor body optimization relatively few, furthermore most ofthe existing control have not considered the torque from mutual inductance. To solve the aboveproblem, based on the analysis of the torque ripple mechanism, and combined with the fastconvergence and global optimization characteristics of the particle swarm algorithm, this paperultimately achieve torque ripple global optimization of the SR motor mainly through improveddesign of the motor body, power topology optimization, and instantaneous torque compensationcontrol.First, the complete structure design of the motor is achieved by the systematic derivationof the SR motor output equation. For the special four-phase8/6pole SR motor, this paper usesthe steady-state finite element method to analyze the differences beween single and two-phaseexcitation mode as well as long and short magnetic circuit in characteristics of flux, inductanceand torque, taking the mutual inductance coupling, saturation effects into account. Theself-inductance and mutual inductance characteristics of SR motor were measured to verify theaccuracy of the finite element simulation. The SR magnetic field-electric circuit-movementcoupling analysis model was established, while time stepping finite element method wasintroduced to study characteristics of the SR motor from start-up to steady-state operating indetail.Second, from the perspective of the motor inherent performance considerations, this paperstudied how to optimize motor performance by design process. The effect of motor pole shapeon the objective functions associated with the optimization variables and constraints wereobtained by finite element analysis. Through the training of generalized regression neuralnetwork, the mathematical models of the objective function were obtained, and adaptiveparticle swarm algorithm was used to solve the optimal design of the SR motor with nonlinear,multi-variable, multi-constraint conditions. The calculation program can be extended to anynumber of phases, and any structure motor, so it can provide a general algorithm for SR motoroptimal design.At the same time, from the view of both motor control and power topology, the doublephases conduction equations including the mutual inductance were derived, and nonlinearsimulation model of the SR motor was carried out using Simulink. The effects of switchingangles and chopping frequency on torque ripple were studied in-depth, so provide a strong basisto choose a reasonable control parameters; the new power topology based on capacitivecompensation was proposed, which improves system power factor and efficiency whilereducing torque ripple.Finally, this paper presents the instantaneous torque control strategy based on generalizedregression neural network with particle swarm optimization, and for the double phasesexcitation mode combined with the torque compensation principle to jointly achieve the the torque ripple optimal control strategy, the method, combined with the torque inverse model, cancompensate the mutual inductance torque due to asymmetric electromagnetic, and torque rippleis significantly reduced.The SRD system experimental platform is built to complete the steady–state experiment,dynamic performance experiment, and the application of the instantaneous torque control andordinary PWM control are used, then the effects of different control modes on phase current,output torque and torque ripple is analyzed. The experimental results show that the stable phasecurrent, robustness, and better anti-interference ability, at all a global optimization control of theSR motor torque ripple is achieved.更多还原