摘要
月面科研站能源系统工况复杂,对其电源系统的稳定性分析十分重要,电源系统建模复杂,将其等效为黑箱子模型,通过输入和输出特性判断稳定性。提出了一种利用动态神经网络根据给定输入估计系统的预测输出,判断系统的稳定性。神经网络模型的训练数据集通过仿真生成,通过选择合适的神经网络结构和超参数,获得最佳的传递函数辨识结果。以Buck电路为例,采用MATLAB的Simulink模块进行仿真得到神经网络训练集,神经网络预测输出能贴近仿真输出,验证了该方法的有效性。
The energy system of the lunar surface research station has complex working conditions,and it is very important to analyse the stability of its power supply system,which is complex to model,and is equated to a black box model to judge the stability through the input and output characteristics.A way was proposed to judge the stability of the system using a dynamic neural network to estimate the predicted output of the system based on the given input.The training dataset of the neural network model was generated by simulation,and the best transfer function identification results were obtained by selecting the appropriate neural network structure and hyperparameters.The neural network training set was obtained by simulation using the Simulink module of MATLAB as an example of a Buck circuit,and the predicted output of the neural network could be close to the simulated output,which verified the effectiveness of the method.
作者
黄宇超
宋相毅
童乔凌
张侨
HUANG Yuchao;SONG Xiangyi;TONG Qiaoling;ZHANG Qiao(School of Automation,Wuhan University of Technology,Wuhan Hubei 430070,China;School of Integrated Circuits,Huazhong University of Science and Technology,Wuhan Hubei 430074,China)
出处
《电源技术》
CAS
北大核心
2024年第3期395-400,共6页
Chinese Journal of Power Sources
