基于LM-BP神经网络的高铁建设环境中短期影响综合评价研究

Research on the Mid-short Term Environmental Impact of High-speed Rail Construction Based on LM-BP Neural Network

根据改进过的BP人工神经网络建立了高铁建设环境中短期影响的三层前馈神经网络综合评价模型。改进过的BP神经网络算法是在传统的算法的基础上使用LM算法,即梯度法和高斯—牛顿迭代法的结合,该种方法提高了收敛速度及精确度。根据既往已经开通运营的铁路项目对生态环境影响的深入调查和研究,建立了高铁建设环境中短期影响评价的指标体系。利用长益城际铁路作为实证,使用LM-BP神经网络模型得到的综合评价分值分别为0.372和0.223,表明该铁路建设对沿线环境的短期和中期影响分别为较小影响和极小影响,评价结果有效。该模型的应用可提高评价结果的客观性和准确性,为高速铁路的选线决策提供科学的评价方法。

In this paper, a three-level forward feedback ANN model of short-term environmental impact of high-speed rail is established based on modified BP ANN. The modified BP ANN is based on traditional LM algorithm, combining gradient method and Gauss-Newton iterative method. The con- vergence speed and accuracy are much improved by this modification. Moreover, through the study and investigation on environmental interference of existing and operating high-speed rail projects, an evaluation model system is constructed. Taking Chang Yi city rail as a sample, the LM-BP ANN mod- el is used to analysis the environmental impact of this construction project. This validation process justifies that the modified BP ANN model is valid, since the model output is an objectively valid and accurate result. The application of this model can provide a scientific evaluation method for the high- speed railway line selection decisions.