Water Quality Forecasting Based On ARIMA And SVM For Xiangjiang Rive

The importance of water quality assessment and water quality forecasting has become increasingly important and has become an significant part of environmental protection management.Water quality forecasting analysis can provide the trend of water quality changes in target basins,thus making it possible for people to take measures in advance to eliminate the occurrence of pollution and achieve the purpose of protecting the overall water quality of the basin.Therefore,water quality prediction has very important practical significance.In this paper,a time series analysis model—ARIMA model combined with Support Vector Regression(SVR)is used to construct a combined forecasting model to predict the water quality parameters of Xiang river.The main work is as follows:(1)Collection and Treatment of water quality parameters in Xiang River Basin.The author sampled once a week for 125 weeks between 2015 and 2017.three water quality parameters were collected from seven cross sections in Luzhou,Yongzhou,and Guiyang,Hengyang,in the Xiangjiang River basin.They were lead(pb),arsenic(As),pH,dissolved oxygen(DO),permanganate index(CODMn),total phosphorus(TP)respectively.For the lack of some water quality parameters,the author complements the missing data with Lagrangian interpolation.Using this method,the data mean is constant and the variance is relatively small,which has less impact on the prediction of time series analysis.(2)Water Quality Prediction Based on ARIMA Model.Using the timing diagram and auto-correlation diagram to judge the stationarity of the model,choose the appropriate d value,that is,the number of variance,convert the non-stationary time series into a stationary time series,and use the relative amount optimal method to achieve the automatic ordering of the ARIMA model.Select ARMA(p,q).When the information of aic,bic,and hqic for all combinations of p and q are less than or equal to 10,the optimal model is the one in which the aic,bic,and hqic information amount is the minimum.(3)Combination forecast based on ARIMA model and SVM.The ARIMA model is a linear model,which has limited ability to predict nonlinear data.Therefore,this paper constructs a combination model of ARIMA and SVM,including parallel combination model and series combination model.The parallel combination model separately calculates the ARIMA and SVM prediction values,and then assigns the prediction values with different weights.After the summation,the parallel model prediction values are obtained;the ARIMA model prediction values are used as the input of the SVM model,and then the SVM model is used to predict the results.The tandem model prediction results are superior to the parallel results.(4)On the basis of the above-mentioned combined forecasting model research,a platform for water quality prediction was established.The platform enables basic functions such as data import,viewing,and forecasting.The platform displays the transformation of water quality data by means of tables and an intuitive line chart,laying the foundation for the future automated monitoring of water quality data platforms.The results of this paper show that the accuracy of the tandem combination forecasting model based on ARIMA and SVM is high,but its effectiveness and universality need to be further researched.