论文部分内容阅读
研究了不同激活函数选取对ＢＰ 网络收敛速度的影响，得出了采用组合激活函数可改善ＢＰ网络的收敛性的结论．以电力变压器的人工神经网络故障诊断为例所选取的ＴＴＳ（Ｔ表示为双曲正切函数，Ｓ表示为Ｓｉｇｍｏｉｄ 函数） 的组合激活函数方式，具有快速收敛性和较高的故障诊断精度．
The influence of different activation functions on the convergence speed of BP network is studied. The conclusion that the combination activation function can improve the convergence of BP network is obtained. Taking the artificial neural network fault diagnosis of power transformer as an example, the combined activation function method of TTS (T represents hyperbolic tangent function and S represents Sigmoid function) is adopted, which has fast convergence and high fault diagnosis accuracy.