ESD模型参数识别的差分进化方法

提出了一种以差分进化算法对静电放电模型进行参数辨识的新方法。以基于Heidler雷电流方程的静电放电模型参数为辨识对象,分别以仿真和实验数据验证了该方法的可行性,并从电流波形整体和局部两方面对拟合效果进行了评估。结果表明,与遗传算法相比,差分进化算法的执行速度更快,所得的辨识参数精度更高,对电流波形的整体和局部关键点的拟合度均高于遗传算法。因此,差分进化算法比遗传算法更适用于解决静电放电模型参数辨识问题。从实例可以看出,差分进化算法不需要过多的初始参数值先验知识,而只需提供一个较宽的初始参数搜索范围即可获得良好的辨识结果。此外,本文还以差分进化算法对Bruce-Golden和Gaussian函数静电放电模型进行参数辨识,验证了该方法的适用性。

Parameter identification of electrostatic discharge model based on differential evolution algorithm

This paper presents a new approach for determining electrostatic discharge(ESD) model parameters by means of differential evolution(DE) algorithm.The parameters of an ESD model based on Heidler equation are selected as the identification objects.The validity of this approach has been confirmed with experimental and simulated current data.The fitting effect has also been evaluated according to the overall and partial sides of the current waveform.Compared with genetic algorithm(GA),the DE approach runs faste...