基于LSTM神经网络的牵引站电气设备耦联体系地震响应预测

Seismic response prediction of electrical equipment interconnected system of traction station based on LSTM neural network

铁路牵引变电站中,软导线-电气设备耦联体系具有较强的几何非线性。为提升系统分析效率,提出一种改进的软导线-电气设备耦联体系地震响应递归预测方法。基于长短期记忆(long short-term memory,LSTM)神经网络与Dropout防止过拟合技术搭建了LSTM神经网络预测模型。建立了充分考虑软导线对相邻设备的耦联作用的软导线-电气设备耦联体系理论分析模型。为验证预测模型的泛化能力,筛选出了41条在峰值、频谱和持续时间上具有较大差异的地震波。并按照递归方案,将选取的地震波以及软导线-电气设备耦联体系理论分析模型计算所得的位移响应,进行滑动切片处理,建立模型输入特征与输出响应标签的映射关系。在此基础上,利用该LSTM神经网络预测模型开展了软导线-电气设备耦联体系设备的地震位移响应预测,并采用多个评价指标进行较为全面的模型性能评估。研究结果表明:LSTM递归预测模型具有良好的地震响应预测性能,搭配Dropout技术能够有效防止模型训练过拟合,提高模型适应能力。对于差异较大的地震波数据,均能够快速预测出误差较小、相关度较高的地震响应,具有较好的准确性、高效性与泛化能力。所提方法能够较高效准确地预测任意时刻的软导线-电气设备耦联体系地震响应,为铁路牵引变电站抗震设计提供新的研究思路。

In the railway traction substation,the flexible conductor-electrical equipment interconnected system has strong geometric nonlinearity.To improve the analysis efficiency of the system,an improved recursive prediction method for the flexible conductor-electrical equipment interconnected system was proposed.The prediction model was established based on recursive long short-term memory(LSTM)neural network and the Dropout regularization.The theoretical analysis model of the interconnected system between the flexible conductors and electrical equipment was established,fully considering the coupling effects of flexible conductors on adjacent equipment.Besides,to fully reflect the generalization ability of the model,41 seismic time histories with large differences in peak ground acceleration(PGA),frequency spectrum,and duration were selected.According to the recursive scheme,the selected seismic time histories,along with the displacement responses obtained by the theoretical analysis model of the flexible conductor-electrical equipment interconnected system,were subjected to sliding-window slicing treatment and established the mapping relationship between model input features and output labels.In addition,the model was used to predict the seismic displacement responses of the interconnected system,and several evaluation indices were used to evaluate the model performance comprehensively.The results indicated that the LSTM recursive prediction model exhibits excellent performance in seismic response prediction.When combined with the Dropout regularization,it effectively prevents model overfitting and improves the adaptability of the model.For seismic time-history data with significant variations,the model can rapidly predict earthquake responses with lower errors and higher correlations,demonstrating high accuracy,efficiency,and generalization capability.This method can quickly and accurately predict the seismic responses of the flexible conductor-electrical equipment interconnected system at any time point,providing a new research idea for the seismic design of railway traction substations.