机器学习驱动的辉光放电等离子体降解碱性紫16性能研究  

Performance study of basic violet 16 degradation by glow discharge electrolysis plasma based on machine learning

作　　者：方野 王玉如 曾静懿 王亚欣 郑伟 李敏睿[1,2] FANG Ye;WANG Yu-ru;ZENG Jing-yi;WANG Ya-xin;ZHENG Wei;LI Min-rui(School of Geography and Tourism,Shaanxi Normal University,Xi'an 710119,China;National Geography Experiment Teaching Demonstration Center,Shaanxi Normal University,Xi'an 710119,China)

机构地区：[1]陕西师范大学地理科学与旅游学院,陕西西安710119 [2]陕西师范大学,地理学国家级实验教学示范中心,陕西西安710119

出　　处：《中国环境科学》2024年第6期3206-3216,共11页China Environmental Science

基　　金：陕西省自然科学基础研究计划-资助项目(2021JM-192)。

摘　　要：为了量化研究辉光放电等离子体(GDEP)降解碱性紫16(BV16)的效率影响因素以进一步提升其降解性能,共收集462条降解实验数据建立数据集,训练并评价了9种回归模型.结果表明,基于梯度提升树(GBDT)的集成学习模型预测性能优异,且以类别型特征提升(CatBoost)算法训练的模型性能最佳(R^(2)=0.988,MAE=2.050%).此外,沙普利加和解释方法(SHAP)对最佳模型的参数影响程度定量解析结果显示,反应时间(43.74%)、初始污染物浓度(23.00%)、氯化钾浓度(15.65%)和平均电流(12.63%)是影响BV16降解效率的关键因素.同时,基于部分依赖图(PDP)提出了参数交互影响优化方案.所建立的CatBoost-SHAP-PDP模型不仅能实现GDEP对BV16降解效果的模拟预测,而且是优化GDEP降解过程变量的有效方法,为GDEP降解染料废水复杂体系领域的建模与应用提供科学依据和技术支持.This study aimed to quantitatively investigate the factors influencing the degradation efficiency of Basic Violet 16(BV16)by Glow Discharge Electrolysis Plasma(GDEP)and to enhance its degradation performance.A dataset was constructed from 462 experimental data points,and 9 regression models were trained and evaluated.The integrated learning models based on the Gradient Boosting Decision Tree(GBDT)demonstrated superior predictive performance,with the model trained using the Categorical Boosting(CatBoost)algorithm exhibiting the highest performance(R^(2)=0.988,MAE=2.050%).The SHapley Additive exPlanation(SHAP)interpretation method was employed to quantitatively analyze the impact of parameters in the optimal model.The quantitative weight ranking results indicated that reaction time(43.74%),initial pollutant concentration(23.00%),KCl concentration(15.65%),and average current(12.63%)were the most significant factors influencing BV16 degradation.Furthermore,Partial Dependence Plot(PDP)analysis was utilized to propose an optimization scheme for parameter interactions.The CatBoost-SHAP-PDP model facilitated the simulation and prediction of BV16 degradation by GDEP and provided an effective method for optimizing the variables in the GDEP process.This research offers a scientific foundation and technical support for modeling and application in the field of complex dye wastewater treatment by GDEP.

关 键 词：辉光放电等离子体 染料废水 SHAP解释方法 机器学习 

分 类 号：X703[环境科学与工程—环境工程]