火电厂用水档案建立及数据分析  

Establishment of water use files and datas analysis in thermal power plants

作　　者：刘克成 张立军 王颖楠 姚鸿淼 LIU Kecheng;ZHANG Lijun;WANG Yingnan;YAO Hongmiao(State Grid Hebei Energy Technology Service Co.,Ltd.,Shijiazhuang 050000,China)

机构地区：[1]国网河北能源技术服务有限公司,石家庄050000

出　　处：《给水排水》2023年第5期74-82,共9页Water & Wastewater Engineering

基　　金：国网河北能源技术服务有限公司课题(TSS2020-13)。

摘　　要：目前我国尚未建立通用的火电厂用水效率评价标准和分级。首先将火电厂用水效率分为优秀、良好、不及格三个等级,选取单位发电取水量、重复利用率、排放水率和废水回用率作为相应的评价指标,结合水平衡测试报告收集了共24组数据,并利用多重插补对缺失数据进行插补处理。接下来运用综合模糊综合评价法、系统聚类法和动态聚类法3种方法的结果对24家火电厂给出合理的评价分级,作为支持向量机的训练样本集。在支持向量机中,针对数据不均衡问题采用G-SMOTE算法进行均衡化处理,有效避免了数据量过少导致插值区域受限的问题。最后利用参数优化后的支持向量机模型对3家火电厂进行用水效率评价,并结合其水平衡报告给出具体的改进建议,为火电厂用水效率评价提供了标准数据集和系统方法。At present,our country hasn't established a general evaluation standard and classification for the water-use efficiency of thermal power plants.This paper firstly divides water-use efficiency into three grades:excellent,good and fail,and choose quantity of water intake for unit power generation quantity,recycle ratio,discharged water ratio and wastewater reuse ratio as corresponding evaluation index.Then 24 sets of data are collected from water balance tests and the missing data are filled in with multiple imputation.Next,comparing the results of fuzzy comprehensive evaluation,systematic clustering and dynamic clustering,the final reasonable classification of the water-use efficiency of 24 thermal power plants is concluded,which is used as the training sample set for support vector machine(SVM).In SVM,G-SMOTE algorithm is used to deal with the imbalanced data,which efficiently avoids the problem of limited interpolation area,caused by insufficient data.At last,this paper uses improved SVM model to evaluate 3 thermal power plants'water-use efficiency and gives specific suggestions for improvement in conjunction with water balance tests.We established standard benchmark and systematic methodology for evaluating water-use efficiency of thermal power plants.

关 键 词：用水效率 多重插补 聚类分析 G-SMOTE 支持向量机 

分 类 号：X703[环境科学与工程—环境工程]