同步电机在线参数辨识的混合算法

根据输出误差法(OEM)、遗传算法和神经网络的优缺点,提出了一种同步电机参数在线辨识的四步法.由于OEM辨识参数需要较好的初值,文中首先用遗传算法在大范围内进行参数寻优,然后以此值作为OEM的初值进行迭代,再用OEM求得的结果训练人工神经网络,最后通过成功训练的神经网络在线辨识各种运行状态下的电机参数.这样使得传统的OEM参数辨识算法、遗传算法和神经网络在辨识参数方面充分地扬长避短,解决了单一算法的不足.在对一台111kVA、440V同步电机进行的仿真试验中,该方法在保证精度的前提下,辨识时间仅为0 008s.

Hybrid Method for On-Line Synchronous Parameter Identification

A method is presented to identify synchronous machine rotor parameters with four steps. Good initial values in parameter identification are necessary for the traditional output error method (OEM) algorithm, so first the genetic algorithms (GA) global searching ability is utilized to find parameters closed to the optimum values; the synchronous machine parameters can be identified accurately with the aid of these parameters; then the BP neural network is exerted a perfect training; and finally a well trained neural network is employed to identify machine parameters under the specific operating conditions. In this way the GA global searching, the OEM local searching ability and the real time capability of BP neural network can fully play their roles.A simulation test for 111 kVA,440 V synchronous machine confirms the validity of this method. It takes only 0.008 s to complete a parameter identification with enough accuracy.