基于人工神经网络和灰色系统的工业废水排放预警方法比较及实用性分析

Comparison and Practical Analysis of Industrial Wastewater Discharge Warning Method Based on Artificial Neural Network and Gray System

为研究遥感图DN值与水质监测数据之间的关系,准确预测水质变化趋势,为环境监管部门监管工业企业提供针对性建议,采用线性回归法、BP神经网络和灰色系统分别建立水质数据与流域遥感图DN值之间的预测模型,对工业企业污水排放进行针对性监管。结果表明,当波段像元亮度值分别作为自变量、水质作为应变量时,线性回归结果并不理想,相关系数R2均在0.8以下,最大为0.7036;BP神经网络训练次数达到45385次时,运算精度已经达到9.99×10-6,MSE为0.0562,训练效果良好;CODMn浓度作为输入数组时,灰色系统模型百分绝对误差为0.6925%,预测效果好。

In order to study the relationship between DN value of remote sensing map and water quality monitoring data,accurately predict the change trend of water quality,and provide targeted suggestions for environmental supervision departments to supervise industrial enterprises,this study adopts linear regression method,BP and gray system to establish a prediction model between water quality data and neural network and gray system to supervise the sewage discharge of industrial enterprises.The experimental results show that when the band image element brightness value respectively as the independent variable,the water quality as the stress variable,the linear return result is not ideal,the correlation coefficient R2 is below 0.80,maximum 0.7036;BP neural network training times reached 45385 times,the operation accuracy has reached 9.99×10-6,MSE is 0.0562,good training effect.When the CODMn concentration acts as an input array,the percent absolute error of the gray system model is 0.6925%,with a good prediction effect.