用于磁化曲线拟合的高精度混合型径向基函数神经网络

NETWORKS IN SIMULATION OF FERROMAGNETIC ELEMENTS' MAJOR HYSTERESIS CURVE

铁磁材料的主磁滞回环是强非线性函数 ,其精确拟合是电力系统暂态仿真中的一个重要课题 ,应用人工神经网络对其进行模拟是一种新尝试。作者针对前馈神经网络的反向传播 BP学习算法收敛速度慢和径向基函数 (RBF)神经网络在拟合中光滑性 (内插和外推能力 )差的缺点 ,提出了一种新型的混合型径向基函数神经网络 ,有效地克服了 BP神经网络和普通径向基函数神经网络在铁磁材料主磁滞回环拟合方面的缺点 ,实际应用获得满意结果。

The major hysteresis loop of ferromagnetic material is a kind of strong non linear function. To precisely fit the major hysteresis loop is an important topic in the power system transient simulation and it is a new attempt to simulate it by artificial neural network (ANN). Because of the low convergence speed of BP (back propagation) learning algorithm for Feed Forward Neural Network and the weak smoothness in the fitting by Radial Basis Function Neural Network, a new hybrid radial basis function neural network is put forward. The above mentioned disadvantages in the fitting of major hysteresis loop are effectively surmounted. The simulation results show that the presented method can fully satisfy the demand of the power system dynamic simulation.