摘要
为探索基于系统辨识的电磁脉冲效应仿真新方法,设计了以阶跃信号和方波脉冲信号为激励源、稳压电源系统为对象的脉冲注入实验,分别采用OE模型和NARX神经网络模型对该系统的脉冲能量耦合传递函数进行建模。结果表明,所建模型均能较好地预测出响应波形,且NARX模型预测能力强于OE模型,两者对阶跃、方波脉冲的预测精度分别达到93.0%、67.4%和76.0%、61.4%以上。两模型的仿真结果证实了系统辨识对电路电磁脉冲响应预测的正确性,为电磁防护设计提供了一种简单有效的仿真新方法。
In order to study the new methods of electromagnetic pulse effects simulation based on system identification, the pulse injection experiment of regulated power supply is designed with step signals and square wave pulse as drive sources. The pulse energy coupling transfer function of circuit system is modeled by OE （Output Error） model and NARX（Nonlinear Autoregressive Network with Exogenous Inputs） Neural Network. As a resuit, both of the models can predict the response wave shapes well, and the predictive ability of the NARX model is better than that of OE model. The goodness - of- fit on the step signal and square wave pulse response wave shapes, predicted by NARX model, is as high as 93.0% and 67.4%, and that predicted by OE model is 76.0% and 61.4%. The simulation results prove the effective application of system identification in electromagnetic pulse response prediction of the circuit. It offers a simple and effective simulating method for electromagnetic protection and design.
出处
《电讯技术》
北大核心
2011年第10期117-121,共5页
Telecommunication Engineering
基金
国家自然科学基金资助项目(50877079)
国防科技重点实验室基金资助项目(9140C87030211JB34)~~
关键词
电磁脉冲效应
电磁防护
系统辨识
最小二乘法
OE模型
NARX神经网络模型
electrorrkagnetie pulse（EMP） effect
electromagnetic protection
system identification
least square algo- rithm
OE model
NARX neural network model
