基于MATLAB BP神经网络的岩溶渗漏污染预测一例

Prediction of pollution of karst leakage in Baiji ardealite site with BP neural network system

本文以摆纪磷石膏堆场为研究对象,采用了地下水物质迁移模型中的"黑箱"模型,即运用MATLAB的BP神经网络建立磷石膏堆场岩溶渗漏污染预测模型,实现了人工神经网络对堆场岩溶渗漏污染的预测。在岩溶渗漏管道为单一管道类型时,模型预测值基本与实测值吻合,误差较小,效果较为理想。但对复杂的岩溶渗漏管道类型,虽然能大致反映出污染物浓度变化的趋势,但模型精度不够,误差较大,因此还需进一步收集数据进行模型的优化,使其达到理想的预测效果。

Karst groundwater pollution problems in Guizhou province are getting more and more prominent. It has become a threat to the groundwater resources of Karst area. Based on evaluating and analyzing the karst groundwater leakage of the slag yard, a model of predicting the Krast groundwater seepage with BP neural network system with NNTOOL of neural network tool box in MATLAB is established and the feasibility of the BP neural networks model in karst groundwater is realized. The predicted value compares well with the measured value based on the Krast groundwater seepage with the type of single channel of small error. Although the model represents the trend of the change with pollutant concentration, the precision is inadequate. Therefore, it needs to reach the ideal prediction with more data.