WEKA环境下模拟预测城市污水处理厂泥饼含水率  

Simulation and Prediction of Moisture Content of Sludge Cake from WWTPs by WEKA

作　　者：李佟[1,2] 李军[1] 付强[2] 

机构地区：[1]北京工业大学北京市水质科学与水环境恢复工程重点实验室,北京100124 [2]北京城市排水集团有限责任公司,北京100022

出　　处：《中国给水排水》2015年第19期76-79,共4页China Water & Wastewater

摘　　要：以7座污水处理厂的20 000多个数据为依据,首先通过灰色关联法确定了对泥饼含水率具有显著影响的指标与输入变量,在此基础上,以WEKA为工具平台,采用M5P分类回归算法对污水厂泥饼含水率指标进行建模,并进行了模拟校核与预测。模拟校核结果表明,WEKA下的M5P算法非常适合于既包含数值型又包含分类型的复合指标组模拟,对于污水厂污泥脱水具有很好的模拟效果。模拟预测结果显示,泥饼含水率预测值和实际值吻合程度较好,平均绝对误差为1.18、均方根误差为1.59。模型可良好地应用于城市污水处理厂泥饼含水率的预测分析,对于污水厂的生产运行具有一定的指导意义。Based on more than 20 000 data collected from 7 WWTPs, the indicators and input variables which have significant effect on moisture content of sludge cake were determined by gray correlation method. On this basis, the moisture content of sludge cake from WWTPs was modeled and simulated u- sing M5P classification and regression algorithms in WEKA platform. The simulation and check results showed that M5P algorithm was well suited for simulation of both numerical and categorical indicators, and it presented excellent simulation effect of dewaterability of sludge from WWTPs. The simulation and prediction results showed that the predicted value of moisture content of sludge cake was consistent with the actual value very well, with average absolute error of 1.18 and the RMS error of 1.59. The model can be well applied to prediction analysis of moisture content of sludge cake from WWTPs, which has a certain significance for the operation of WWTPs.

关 键 词：泥饼 含水率 数据挖掘 WEKA M5P算法 

分 类 号：X703[环境科学与工程—环境工程]