神经网络模型在混凝效果预测中的应用现状及展望  

The application status and prospects of neural network models in coagulation effect prediction

作　　者：曹昕[1] 马拓 杨俊坡 杨耿 郑兴 CAO Xin;MA Tuo;YANG Junpo;YANG Geng;ZHENG Xing(Faculty of Water Resources and Hydroelectric Engineering,Xi’an University of Technology,Xi’an 710048;Guanxing Intelligent Technology(Xi'an)Co.,Ltd.,Xi’an 710054;School of Electronic Information and Artificial Intelligence,Shaanxi University of Science and Technology,Xi’an 710021;College of Information and Control Engineering,Xi’an University of Architecture and Technology,Xi’an 710055)

机构地区：[1]西安理工大学水利水电学院,西安710048 [2]观星智能科技(西安)有限责任公司,西安710054 [3]陕西科技大学电子信息与人工智能学院,西安710021 [4]西安建筑科技大学信控学院,西安710055

出　　处：《环境科学学报》2023年第12期186-193,共8页Acta Scientiae Circumstantiae

基　　金：陕西省科技厅青年创新团队科研计划项目(No.22JP054)。

摘　　要：受原水水质、水力条件、混凝剂种类及混凝剂投加调控机制等诸多因素影响,混凝处理效果与混凝剂投加量之间往往呈现复杂的非线性关系.传统研究着眼于混凝机理、混凝剂特性及混凝过程水质、絮体信息采集与投药调控等方面,尚未能构建基于全混凝操作流程的、用以实现混凝效果预测的普适性理论.神经网络模型因其具有强大的学习能力,近年来在混凝效果预测研究领域中受到了广泛的关注.本文通过对混凝效果预测算法研究历史进行回顾与分析,总结神经网络模型在混凝效果预测中的研究现状和脉络,深入分析不同数据来源与数据格式的优缺点,从实验装置、水质参数、投加控制和数据时序性、混凝剂构效研究与产品技术开发等方面,展望神经网络模型在混凝效果预测中的未来研究方向.The effect of coagulation is affected by many factors such as raw water quality,hydraulic conditions,type of coagulant and the mechanism of coagulant dosage control.There is often a complex nonlinear relationship between the performance of coagulation treatment and coagulant dosage.Previous studies have focused on coagulation mechanisms,coagulant characteristics,water quality in coagulation process,the collection of flocculant information and dosing monitoring etc..However a general theory based on the whole process of operation to predict its performance is still not available.Neural network models have attracted extensive attention in the field of coagulation during recent years due to their strong learning capabilities.In this paper,the previous studies on the prediction of coagulation performance using various algorithms including neural network models are reviewed and the advantages and disadvantages of both different data sources and data formats are thoroughly analyzed.Furthermore,the research direction of neural network models in the prediction of coagulation performance in future is prospected from the aspects of experimental equipment,water quality,dosage control and data timing,coagulant’s chemical structure-function research and product development.

关 键 词：深度学习 混凝效果 神经网络模型 智能Jar test 水质参数 随机森林模型 

分 类 号：X32[环境科学与工程—环境工程] X703