基于神经网络转速估计的异步电机矢量控制系统研究

发布时间：2018-01-15 20:39

本文关键词：基于神经网络转速估计的异步电机矢量控制系统研究　出处：《湖北工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：传统交流调速的难题一直阻碍着交流异步电机的发展,而矢量控制技术可以使交流电机获得直流电机的控制特性。为了实现高速高精度调速要求,转速闭环控制是核心,故高效智能的控制算法是矢量控制系统的性能好坏的关键。由于编码器等速度传感器在恶劣环境下存在精度低、稳定性差等问题,利用智能算法对电机转速进行实时估计,从而实现基于矢量控制技术的无速度传感器电机控制系统是国内外交流异步电机控制研究的热点所在。本文在深入研究矢量控制技术基础上,首先对控制系统的闭环控制算法进行改进,之后取代传统编码器,转而利用智能算法估计,针对无速度传感器矢量控制系统展开研究。取得的成果如下。(1)基于矢量控制技术实现解耦控制。本文被控对象为三相交流异步电机,首先分析其数学模型,针对其数学模型非线性、强耦合等数学特性导致调速难的问题,在分析了矢量控制技术原理之后,利用其实现对定子电流的解耦,实现励磁与转矩分量解耦控制,从而获得直流电机的控制特性。(2)利用模糊控制算法改进传统采用PID算法的转速控制器,实现高速高精度矢量控制系统。在控制系统中,结合模糊控制与PID控制的特点,采用模糊PID控制算法实现矢量控制系统闭环控制,从而解决了传统PID在复杂情况下无法实现高速高精度控制性能的问题,并基于华中数控五轴数控机床的主轴控制系统进行了相关实验,应用传统PID控制器与模糊PID控制器进行对比实验,实验结果表明模糊PID在响应速度、控制超调及稳态误差方面较传统PID算法有显著的优势。(3)利用神经网络算法估计转速,实现基于矢量控制的无速度传感器高速高精度控制系统。针对矢量控制系统中,传统传感器成本高、使用环境有限等问题,采用BP神经网络对转速进行估计,设计了神经网络转速估计模型与相应学习算法。并针对BP神经网络算法在转速估计方面存在的问题,提出基于动量法与启发式预处理法改进的神经网络算法。利用Maltab搭建系统模型仿真,对工作性能进行验证。证明了BP神经网络估计的转速可以很好的跟踪实际转速,改进BP神经网络离线训练速度更快更稳定,工作过程中具有更高的精度。
[Abstract]:The traditional AC speed regulation problem has been hampering the development of AC asynchronous motor, and vector control technology can make AC motor obtain the control characteristics of DC motor, in order to achieve high speed and high precision speed regulation requirements. Closed-loop speed control is the core, so efficient and intelligent control algorithm is the key to the performance of vector control system. The intelligent algorithm is used to estimate the motor speed in real time. Therefore, the speed sensorless motor control system based on vector control technology is a hot spot in the research of AC asynchronous motor control at home and abroad. Firstly, the closed-loop control algorithm of the control system is improved, then instead of the traditional encoder, the intelligent algorithm is used to estimate. The research of speed sensorless vector control system is carried out. The results are as follows: 1) decoupling control based on vector control technology. The object of this paper is three-phase AC asynchronous motor. Firstly, this paper analyzes its mathematical model, aiming at the difficulty of speed regulation caused by nonlinear and strong coupling of its mathematical model, after analyzing the principle of vector control technology, it realizes decoupling of stator current. The control characteristics of DC motor are obtained by decoupling the excitation and torque components. The fuzzy control algorithm is used to improve the traditional speed controller using PID algorithm. In the control system, combined with the characteristics of fuzzy control and PID control, the closed loop control of vector control system is realized by using fuzzy PID control algorithm. Thus solving the problem that traditional PID can not achieve high speed and high precision control performance under complex circumstances, and based on the spindle control system of Huazhong CNC five-axis CNC machine tool, the related experiments are carried out. The comparison between the traditional PID controller and the fuzzy PID controller is carried out. The experimental results show that the response speed of fuzzy PID is high. Control overshoot and steady-state error has a significant advantage over the traditional PID algorithm.) the neural network algorithm is used to estimate the rotational speed. The speed sensorless high speed and high precision control system based on vector control is realized. BP neural network is used to estimate the speed of the vector control system in view of the problems of high cost and limited use environment of the traditional sensor in the vector control system. The neural network speed estimation model and the corresponding learning algorithm are designed, and the BP neural network algorithm in speed estimation problems. An improved neural network algorithm based on momentum method and heuristic preprocessing method is proposed, and the system model simulation is built with Maltab. It is proved that the estimated speed of BP neural network can track the actual speed very well, and the speed of off-line training of improved BP neural network is faster and more stable, and the working process has higher precision.
【学位授予单位】：湖北工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM343;TP273

【参考文献】