水质指标具有多元相关性、时序性和非线性的特点，为有效预测河流水质变化，针对水质数据存在缺失和异常的问题，提出基于灰色关联分析-门控循环单元(Grey Relational Analysis-Gated Recurrent Unit，GRA-GRU)的水质预测模型。以淮河流域水质数据为样本，使用线性插值修补缺失数据和剔除的异常数据。使用灰色关联分析计算不同水质指标间的相关性，选择高相关性的水质指标以确定输入变量，并使用门控循环单元(Gated Recurrent Unit，GRU)预测不同的水质指标。将GRA-GRU的预测结果与反向传播神经网络(Back Propagation Neural Network，BPNN)、循环神经网络(Recurrent Neural Network，RNN)、长短期记忆神经网络(Long Short Term Memory，LSTM)、GRU 及灰色关联分析-长短期记忆神经网络(Grey Relational Analysis-Long Short Term Memory，GRA-LSTM)进行对比分析，结果显示GRA-GRU在不同水质指标预测上具有较好的适应性，可以有效降低预测误差。其中，与其他模型相比，GRA-GRU预测的化学需氧量在均方根误差上分别降低了 3。617％、0。681％、0。478％、1。505％和 0。471％。
Water quality prediction model of Huaihe River Basin based on GRA-GRU
The paper proposes a water quality prediction model based on grey relational analysis and gated recurrent unit(GRA-GRU)to solve the issues of data missing and abnormal data in water quality data,as well as the multiple correlation,timing,and nonlinearity of water quality indicators.Two experiments are carried out to verify the predict performance of the water quality prediction model with the water quality data set of Huaihe river basin as a case study sample.The first experiment focuses on data processing,including the processing of missing data and abnormal data of water quality data,and the correlation analysis of water quality indicators.Linear interpolation is adopted to process the missing data,while a combination of boxplot and histogram is adopted to analyze and eliminate the abnormal data,which is then repaired the abnormal data with linear interpolation.The grey relational analysis is adopted to calculate the correlation between different water quality indicators,and water quality parameters with high correlation are retained to determine the inputvariables of the water quality prediction model.The data processing results demonstrate that repairs can be made using linear interpolation without altering the pattern of data change.After the abnormal data is eliminated,the distribution of the water quality data is more centralized.The calculation efficiency and prediction accuracy can both be improved in the water quality prediction model by using the grey relational analysis to reduce the quantity of data it needs as input.A water quality prediction model is created using a gated recurrent unit network to predict various water quality indicators based on the results of the first experiment.The water quality prediction model learning based on the water quality training data set is carried out,including training times,model training batches,and the structure of the gated recurrent unit network,to optimize the water quality prediction model.The second experiment compares the prediction results of the water quality prediction model to BP,RNN,LSTM,GRU and GRA-LSTM based on the water quality indexes of pH,dissolved oxygen,chemical oxygen demand,NH3-N,total phosphorus,electrical conductivity,turbidity,and total nitrogen.Comparative experimental results show that the mean square error of chemical oxygen demand decreased by 3.617％,0.681％,0.478％,1.505％,and 0.471％,respectively,while the mean absolute error decreased by 10.392％,2.301％,4.047％,5.111％and 0.454％,respectively.The coefficient of determination also increased by 1.762％,0.574％,0.495％,0.579％,and 0.392％,respectively.The results of the experiment show that the water quality prediction model proposed in this paper can effectively reduce the prediction error and improve the fitting degree between the predicted result and the actualvalue,having better adaptability to different water quality indicators.
NETL
NSTL
万方数据
维普
