摘要
研究异步电机无速度传感器辩识问题。在电机无速度传感器辩识过程中,为保证电机系统的实时调节的稳定性和准确性,传统的BP神经网络存在网络结构难以确定,极易陷入局部最优解,导致转速辩识慢,精度低的难题。为了提高电机速度辩识准确率,提出一种粒子群和BP神经网络算法相结合的转速辩识方法。采用粒子群来优化BP神经网络粒的权值和阈值,将粒子群算法全局搜索能力和BP算法的局部寻优特点的互补,以提高BP神经网络的收敛速度及精度,将优化后神经网络转速辩识器用于直接转矩控制系统中。在Matlab平台上进行了无速度传感器控制系统的建模仿真。仿真结果表明,该算法加快了辩识速度,提高了转速的辩识精度,具有良好辩识效果。
Asynchronous motor speed identification of speed-sensorless is studied. The traditional BP neural net- work is difficult to determine the network structure and has local optimal problem, leading to slow speed and be of low accuracy. In order to improve the speed of identify effect, this paper puts forward a speed identify method based parti- cle swarm algorithm and the BP neural network. This method adopts the particle swarm to optimize the weights and threshold of BP neural network, and realizes the particle swarm algorithm global search capability and BP algorithm local situation. To improve the BP neural network convergence speed and accuracy, the optimized neural network is used to control speed identifier. Simulation experiment is carried out on the platform of Matlab/Simulink. Simulation results show that this algorithm has accelerated the identify speed and improved the identify accuracy, and therefore this algorithm has good identify effect.
出处
《计算机仿真》
CSCD
北大核心
2011年第9期238-241,共4页
Computer Simulation