基于误差反向传播神经网络的GSM-R场强预测  

Error Back Propagation Neural Network-based Prediction of GSM-R Field Strength

作　　者：丁珣 DING Xun

机构地区：[1]中国铁建电气化局集团有限公司

出　　处：《电气化铁道》2022年第1期67-70,共4页Electric Railway

摘　　要：场强预测是评估和衡量GSM-R系统通信质量的一种重要手段,准确高效的场强预测对信道建模提出了更高的要求。本文提出利用人工神经网络中应用最广泛的误差反向传播(Back Propagation,BP)神经网络预测GSM-R系统的场强。预测结果表明:无需进行三维场景建模,利用表征非线性关系的BP网络即可实现准确的场强预测;相比于无环境特征的数据集,携带环境特征的数据集生成的预测模型精度较高、误差较低。场强预测有助于研究人员理解无线通信特性,更好地服务于GSM-R系统的规划与设计。Field strength prediction is an important means to evaluate and measure the communication quality of GSM-R system.Accurate and efficient field strength prediction puts forward higher requirements for channel modeling.The paper proposes to predict the field strength of GSM-R system by using the back propagation(BP)neural network,which is widely used in artificial neural network.The prediction results show that the accurate field strength prediction can be realized by using BP network representing nonlinear relationship without 3D scene modeling;compared with the data set without environmental characteristics,the prediction model generated by the data set with environmental characteristics has higher accuracy and lower error.Field strength prediction is helpful for researchers to understand the characteristics of wireless communication and better serve the planning and design of GSM-R system.

关 键 词：场强预测 GSM-R系统 人工神经网络 信道建模 

分 类 号：U285.4[交通运输工程—交通信息工程及控制]