蚁群BP网络转速辨识器的优化  

Optimization of Speed Identification Based on BP Neural Network of Ant Colony Algorithm

作　　者：曹承志[1] 王伊凡[1] 

机构地区：[1]沈阳工业大学信息科学与工程学院,辽宁沈阳110023

出　　处：《电机与控制应用》2008年第12期5-8,26,共5页Electric machines & control application

基　　金：辽宁省自然科学基金资助项目(20032032);教育部"春晖计划"合作科研项目(Z2005-2-11008);辽宁省教育厅高校科研项目(20206331)

摘　　要：将排序加权的方法引入基本蚁群算法中,用改进型蚁群算法优化BP神经网络的权值和阈值,有效地解决了BP神经网络训练时容易陷入极小值的缺点,提高了收敛速度,得到了一种时间效率和求解效率都比较好的启发式方法,即改进型蚁群神经网络。运用该方法对直接转矩控制系统中的电机转速进行了辨识。仿真试验结果表明:该改进型蚁群神经网络不仅具有广泛的映射能力,还明显提高了运算效率,能够准确地辨识电机转速,具有良好的辨识效果,实现了无速度传感器直接转矩控制。A rank-weight-based version was added into ant colony optimization to form a modified ant colony optimization algorithm, The weights and thresholds value of neural networks were optimized by using this new algorithm. This method can solve the disadvantage that run into partial minimum value of NN, at the same time it can improve the convergence rate, a heuristic approach which goods at time efficiency and derivation efficiency was propsed, that＇ s ACOrw-BP. The method was used in the .speed identification of motors in direct toque control （DTC）. The results of simulation show that the new ant colony neural network not only is able to map widespread , but also to enhance the operation effectiveness. The speed of motor can be accurately identified by using this method and the result is good, and implement the direct toque control of sensorless.

关 键 词：蚁群算法 神经网络 直接转矩控制 转速辨识 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP273[自动化与计算机技术—控制科学与工程]