基于XGBoost算法构建城镇污水处理厂进水BOD_(5)水质预测模型  

Influent BOD_(5) Prediciton Model Based on XGBoost Algorithm for Urban Wastewater Treatment Plant

作　　者：金秋怡 杨溢 刘宝峰 蔡碧婧 卢颂宇 JIN Qiuyi;YANG Yi;LIU Baofeng;CAI Bijing;LOU Songyu(SUEZ Asia Water Operation,Shanghai 200000,China)

机构地区：[1]苏伊士亚洲水务运营,上海200000

出　　处：《广州化工》2025年第6期153-157,共5页GuangZhou Chemical Industry

摘　　要：为解决污水处理厂进水生化需氧量(BOD_(5))不易检测,滞后性强的问题,以某市政污水处理厂2012年1月至2023年3月共4039组每日进水水质数据,构建了极限梯度提升(XGBoost)进水BOD_(5)预测模型,并与线性回归(LR)、支持向量机(SVR)模型以及人工神经网络(MLP-NN)模型进行对比,验证其可靠性、实用性和先进性。结果表明,XGBoost模型训练过程R~2达到0.982,模型测试的R~2为0.875,平均绝对误差(MAE)为22.376,平均绝对偏差率(MAPE)为11.3%,其性能参数均优于其他模型。XGBoost模型测试的平均相对百分偏差(RP)为5.52%,符合HJ505-2009《水质五日生化需氧量(BOD_(5))的测定稀释与接种法》对BOD_(5)检测的相对偏差要求。该模型表现出较高的准确率和良好的预测优势,为实现污水处理厂进水BOD_(5)的实时预测提供了有效的手段。In order to solve the problem that the biochemical oxygen demand(BOD_(5))of the influent water of the sewage treatment plant is not easy to detect and has strong lag,the Extreme Gradient Enhancement(XGBoost)influent BOD_(5) prediction model based on a total of 4039 sets of daily influent water quality data from January 2012 to March 2023 of a municipal sewage plant was constructed,and compared with the linear regression(LR),support vector machine(SVR)model and artificial neural network(MLP-NN)model to verify its reliability,practicability and advancement.The results showed that the R2 of XGBoost model training process reached 0.982,R2 tested by the model was 0.875,the average absolute error(MAE)was 22.376,and the average absolute deviation rate(MAPE)was 11.3%,and its performance parameters were better than those of other models.The average relative percentage deviation(RP)of XGBoost model test was 5.52%,which met the relative deviation requirements of HJ505-2009 Dilution and Inoculation Method for the Determination of Five-day Biochemical Oxygen Demand(BOD_(5))in Water Quality.The model showed high accuracy and good prediction advantages,which provided an effective means to realize the real-time prediction of BOD_(5) influent water in sewage plants.

关 键 词：市政污水 BOD_(5)预测 XGBoost AI算法 

分 类 号：X703[环境科学与工程—环境工程]