The Research Of Velocity Observer Based On BP Neural Network Optimized By Genetic Algorithm

With the rapid development of power electronics technology, various types of motor control strategy have been constantly evolving. Direct Torque Control technology is a new type of high-performance control of AC variable speed drive technology,which is developed after Vector Control. By calculating and controlling the AC motor torque in the stator coordinate,direct torque control can switch the state of inverter directly in a optimal way. This control method avoided the complex calculation,with strong robustness,good dynamic and static performance, facilitate digital implementation, has already got more and more attention. But its poor low speed performance, the installation of speed sensor will bring a series of shortcomings: Costs more, inconvenience of installation and maintenance difficulties. Speed sensorless control system detects the voltage of stator,current,so as to estimate the speed,replacing the traditional speed sensor,improving the reliability of the system.So that speed sensorless has a good application prospects.This article described the basic principles of direct torque control,analyzed the direct torque control system control program in detail,and elaborated the regulation of flux and torque. I have studied the strategy of the Speed sensorless control, Pointed out in BP neural network learning algorithm,because of the value of the weights and threshold are given in a random way, so that it is easy to fall into local optimum. As a widely used optimization algorithm, GA combined with BP neural network can avoid the defects caused by the randomness.Base on this theory I proposed a strategy that genetic algorithm combined with BP neural network. In the experiment platform, I conducted a data collection experiment. Undering the MATLAB environment, I wrote a program and conduct a simulation experiment.In the experiment platform, the simulation result shows that the application of the Intelligent algorithm using in the neural networks gets much more optimized, Intelligent algorithm based on speed sensorless direct torque control system has a more favorable stability characteristics.