高级检索+

基于三种机器学习模型的太湖总氮浓度预测

Prediction of Nitrogen Concentration in Taihu Lake Based on AdaBoost Machine Learning Model

  • 摘要: 市政污水处理厂出水是自然水体的重要污染来源,处理后的生活污水排入水体,对受纳水体水质产生重要的影响。为实现对太湖总氮浓度的准确预测,收集并分析了太湖水体水质监测数据和太湖流域212个污水处理厂的实时运行监测数据,采用皮尔逊相关系数(Pearson correlation coefficient)分析了太湖水体总氮浓度与市政污水处理厂相关运行指标的相关性,结合相关性较高的前五项指标与太湖水质监测数据,利用临近算法(KNN),决策树以及AdaBoost三种机器学习模型对太湖水体总氮浓度月平均值进行了预测。其中AdaBoost的精度更高、准确性更好,拟合优度为0.84,平均绝对误差在14.08%以内。模型特征重要性分析表明,太湖硝态氮,氨氮和总磷等指标对总氮浓度预测有重要的影响。

     

    Abstract: Domestic waste water is an important source of natural water pollution. The treated domestic waste water is discharged into natural water bodies through the pipe network,which has a serious impact on the quality of the receiving water body. The water quality monitoring data from Taihu Lake and the plant-resolved and actual operation data of 212 waste water treatment plants in Taihu Lake from 2007 to 2015 are collected and analyzed in order to accurately predict the total nitrogen in Taihu Lake. The Pearson correlation coefficient is used to calculate the relationship between Taihu Lake water quality and the WWTP effluent indexes. For the top five items with high correlation,three machine learning models,K-Nearest Neighbors(KNN),Decision Tree,and AdaBoost are used to predict the monthly average TN in Taihu Lake. In general,AdaBoost has higher precision and better accuracy,with a goodness-of-fit index of 0.84 and a root mean square error less than 14.08%,which establishes a good mathematical model for predicting of TN concentration of Taihu Lake. Meanwhile,the model finds that NO3-N,NH4-N,TP in Taihu Lake and NH4-N in effluent of WWTP can cause an important impact on the TN concentration of Taihu Lake.

     

/

返回文章
返回