改进Elman神经网络模型在地铁沉降监测中的应用  

Application of Improved Elman Neural Network Model in Subway Settlement Monitoring

作　　者：徐超良 周波 XU Chaoliang;ZHOU Bo(Ningbo Almaty Digital Technology Co.,Ltd.,Ningbo 315000,China)

机构地区：[1]宁波市阿拉图数字科技有限公司,浙江宁波315000

出　　处：《测绘与空间地理信息》2024年第12期207-210,共4页Geomatics & Spatial Information Technology

摘　　要：以某地铁沉降监测数据为例,提出一种改进Elman神经网络预测模型。首先,发挥局部均值分解(LMD)在信号自适应分解的优势,使用该算法对地铁沉降监测序列进行多尺度分解,得到具有不同尺度特征的乘积函数(PF);其次,发挥Elman神经网络模型在数据序列预测中的优势,对不同PF分量进行训练与预测;最后,重构不同PF分量预测结果得到最终预测成果。实验表明,本文提出的组合预测模型较单一的BP神经网络模型、Elman神经网络模型的预测精度更高,其中均方根误差(RMSE)分别降低了1.0602 mm、0.0698 mm;平均绝对误差(MAE)分别降低了0.8660 mm、0.0474 mm;平均绝对误差百分比(MAPE)分别降低了0.2189、0.0068。This paper proposes an improved Elman neural network prediction model based on a subway settlement monitoring data.First,it takes the advantage of local mean decomposition(LMD)in signal adaptive decomposition,and uses this algorithm to decompose the subway settlement monitoring sequence at multiple scales to obtain the product function(PF)with different scale characteristics;secondly,it takes the advantage of Elman neural network model in data series prediction to train and predict different PF components;finally,the final prediction result is obtained by reconstructing the prediction results of different PF components.The experiments and results show that the combination prediction model proposed in this paper has higher prediction accuracy than the single BP neural network model and Elman neural network model,in which the root mean square error(RMSE)is reduced by 1.0602 mm and 0.0698 mm respectively;the mean absolute error(MAE)decreased by 0.8660 mm and 0.0474 mm respectively;the mean absolute error percentage(MAPE)decreased by 0.2189 and 0.0068 respectively.

关 键 词：局部均值分解 ELMAN神经网络 组合模型 地铁沉降预测 精度分析 

分 类 号：P25[天文地球—测绘科学与技术] TB22[天文地球—大地测量学与测量工程]