基于马尔科夫链改进灰色神经网络的水质预测模型

根据汾河运城段的实际情况,应用改进灰色神经网络对水质进行预测.在数据处理以及关联度分析的基础上,选取关联度较高的氨氮、挥发酚、水温、BOD5及COD作为灰色神经网络的输入节点.应用灰色神经网络对水质进行预测,再用马尔科夫修正误差残值,可使修正值更加接近实测值.灰色神经网络的相对误差为68.44%～4.69%,改进灰色神经网络将相对误差为41.96%～2.23%,可见改进神经网络的预测精度更高.改进灰色神经网络模型,结合了灰色神经网络和马尔科夫的优点,提高了预测的精度,并以汾河河津大桥监测断面的水质预测为例,验证了该方法的可行性.

Water quality prediction model based on Markov chain improving gray neural network

The improved gray neural network is applied to predict water quality according to actual situation of Fenhe river.On the basis of data processing and correlation analysis,NH3-N,volatile phenol,water temperature,BOD5 and COD are chosen as the input nodes of gray neural network model.The gray neural network is applied to predict water quality and then Markov chain is used to modify the residual series for the sample of bigger error.The correction result is close to the measured value.Relative error is 68.44%4.69% based on gray neural network.Relative error was 41.96%-2.23% based on improved gray neural network,the correction result is close to the measured value.Improved gray neural network model,combined the advantages of gray neural network and Markov chain.The results indicate that the method can improve the prediction accuracy.This prediction model was identified by taking water quality prediction into the Hejin bridge monitoring section.