The Research On Vector Control Of Induction Motor

The mathematical model of AC motor is high-order, strong coupling, multi-variable nonlinear systems, the most effective control of it is the vector control first raised in 1970s. Since the vector control proposed, it has significantly improved the dynamic performance of AC drive system, which is sure to be a qualitative leap of AC drive technology.In modern AC drive system, in order to achieve high precision closed-loop speed control, general motors use sensors such as photoelectric encoder to measure the rotor speed. However, the installation of coding device brings shortcomings as the following: the cost of the system greatly increased; reduce the simple and solid characteristics of asynchronous motor; increase the moment of inertia of the system. Moreover, on some occasions it's impossible to install speed sensors. With speed sensorless vector control, we just need to measure the tip voltage and current, then use them to speculate rotor flux and speed, to achieve the vector control of AC motor, while eliminating the cost and reliability problems caused by sensor, it has become a research focus of AC drive technology.Artificial neural network is composed of interconnected artificial neural cell, which has the capacity of self-learning, adaptive, parallel processing of information, so it has been widely used in many fields, such as pattern recognition, automation and signal processing, and so on. From the research results already made, we can see the future of artificial neural network as energetic. But there are still some problems, such as the intelligence level is not quite high, some theoretical issues about the network analysis and synthesis (stability, convergence) have not been satisfactorily resolved yet. However, as people's understanding about the information processing mechanism of brain goes deeper, and the improvement of network intelligence, the artificial neural network is sure to play key roles in more fields.This paper will establish the vector control model of traction motor with sensor, do some simulation research work on it. Based on the comparison of several speed-estimated projects, the paper bring neural network to build model to estimate the rotor speed, to achieve the goal of sensorless vector control. RBF and BP networks are taken to build model using the sample input respectively, based on the comparison and analysis, conclusion is made about the feasibility of each network. Finally the article discusses the necessity and feasibility of neural network hardware implementation under existing conditions, gives a primary block diagram design of the hardware implementation.