基于STA-K均值聚类的电化学废水处理过程离子浓度软测量

doi:10.11949/0438-1157.20190279

摘要/Abstract

摘要：

针对电化学废水处理过程出口离子浓度无法在线检测的问题，提出了一种基于状态转移的K均值聚类算法的软测量建模方法。在分析内部反应机理的基础上，结合物料平衡和吸附动力学定理建立电化学过程的机理模型；由于单一的软测量模型难以满足实际的精度要求，提出一种基于状态转移的K均值聚类算法将原始数据集进行聚类，应用状态转移算法对K均值算法的初始聚类中心进行优化，同时，引入离群值矩阵动态迭代同时实现数据聚类和异常值检测；最后，对聚类后的不同训练子集分别建立子模型，综合各子模型得到基于多模型切换方法的软测量模型。通过某废水处理厂的现场数据进行实例验证，结果证明了所建立的电化学废水处理过程离子浓度软测量模型合理有效。

关键词: 软测量, 废水处理过程, 动力学模型, 状态转移算法, 电化学过程

Abstract:

Aiming at the problem that the ion concentration cannot be detected online during the electrochemical wastewater treatment process, a soft-sensor modeling method based on K-means clustering algorithm of state transition is proposed. On the basis of analyzing the mechanism in the electrochemical process, the mechanism model of the electrochemical process was established according to the material balance and adsorption kinetics. Since single model cannot meet the requirements of accuracy, a K-means clustering algorithm based on state transition is proposed to cluster the original data set. This algorithm uses the state transfer algorithm to optimize the initial clustering center of the K-means algorithm,and introduce a matrix to realize clustering and outlier detection simultaneously during iterative process. Then,after clustering, the sub-models are established with the training subsets respectively, and the soft measurement model based on the multi-model switching method is obtained by synthesizing the sub-models.Finally,via field data verification of a wastewater treatment plant, the results show that the soft measurement model of ion concentration in electrochemical wastewater treatment process is reasonable and effective.

Key words: soft measurement, wastewater treatment process, kinetic modeling, state transition algorithm, electrochemical process

中图分类号:

TP 274

秦美华, 朱红求, 李勇刚, 陈俊名, 张凤雪, 李文婷. 基于STA-K均值聚类的电化学废水处理过程离子浓度软测量[J]. 化工学报, 2019, 70(9): 3458-3464.

Meihua QIN, Hongqiu ZHU, Yonggang LI, Junming CHEN, Fengxue ZHANG, Wenting LI. Soft-sensor method for ion concentration of electrochemical wastewater treatment based on STA-K-means clustering[J]. CIESC Journal, 2019, 70(9): 3458-3464.

图/表 4

参考文献 30

1	Kabdaşlı I , Arslan-Alaton I , Ölmez-Hancı T , et al . Electrocoagulation applications for industrial wastewaters: a critical review[J]. Environmental Technology Reviews, 2012, 1(1): 2-45.
2	Moussa D T , El-Naas M H , Nasser M , et al . A comprehensive review of electrocoagulation for water treatment: potentials and challenges [J]. Journal of Environmental Management, 2017, 186(Pt 1): 24-41.
3	Al-Shannag M , Al-Qodah Z , Bani-Melhem K , et al . Heavy metal ions removal from metal plating wastewater using electrocoagulation: kinetic study and process performance[J]. Chemical Engineering Journal, 2015, 260: 749-756.
4	曾毅夫, 刘君, 邱敬贤, 等 .电化学法处理含重金属电镀废水研究[J]. 再生资源与循环经济, 2018, 11(6): 42-44.
	Zeng Y F , Liu J , Qiu J X , et al . Electrochemical treatment of electroplating wastewater containing heavy metals [J]. Renewable Resources and Recycling Economy, 2018, 11(6): 42-44.
5	曹鹏飞, 罗雄麟 .化工过程软测量建模方法研究进展[J]. 化工学报, 2013, 64(3): 788-800.
	Cao P F , Luo X L . Research progress of soft sensing modeling method for chemical process [J]. CIESC Journal, 2013, 64(3): 788-800.
6	Balasubramanian N , Kojima T , Srinivasakannan C . Arsenic removal through electrocoagulation: kinetic and statistical modeling [J]. Chemical Engineering Journal, 2009, 155(1/2): 76-82.
7	Vasudevan S , Lakshmi J , Sozhan G . Studies on the Al-Zn-In-alloy as anode material for the removal of chromium from drinking water in electrocoagulation process [J]. Desalination, 2011, 275(1/2/3): 260-268.
8	Hu C Y , Lo S L , Kuan W H . Simulation the kinetics of fluoride removal by electrocoagulation (EC) process using aluminum electrodes[J]. Journal of Hazardous Materials, 2007, 145(1): 180-185.
9	Hakizimana J N , Gourich B , Chafi M , et al . Electrocoagulation process in water treatment: a review of electrocoagulation modeling approaches [J]. Desalination, 2017, 404: 1-21.
10	Sasson M B , Calmano W , Adin A . Iron-oxidation processes in an electroflocculation (electrocoagulation) cell [J]. Journal of Hazardous Materials, 2009, 171(1/2/3): 704-709.
11	赵斐, 陆宁云, 杨毅 . 基于工况识别的注塑过程产品质量预测方法[J]. 化工学报, 2013, 64(7): 2526-2534.
	Zhao F , Lu N Y , Yang Y . Product quality prediction method for injection molding process based on condition identification [J]. CIESC Journal, 2013, 64(7): 2526-2534.
12	陈定三, 杨慧中 . 用于多模型软测量的扩张搜索聚类算法[J]. 计算机与应用化学, 2011, 28(4): 407-410.
	Chen D S , Yang H Z . Extended search clustering algorithm for multi-model soft sensing [J]. Computer & Applied Chemistry, 2011, 28(4): 407-410.
13	丛秋梅, 张北伟, 苑明哲 . 基于同步聚类的污水水质混合在线软测量方法[J]. 计算机工程与应用, 2015, 51(24): 27-33+66.
	Cong Q M , Zhang B W , Yuan M Z . Online soft measurement method for mixed wastewater quality based on synchronous clustering [J]. Computer Engineering and Application, 2015, 51(24): 27-33+66.
14	余伟, 罗飞, 杨红, 等 . 基于多神经网络的污水氨氮预测模型[J]. 华南理工大学学报(自然科学版), 2010, 38(12): 79-83.
	Yu W , Luo F , Yang H , et al . Prediction model of ammonia nitrogen in sewage based on multi-neural network [J]. Journal of South China University of Technology (Natural Science Edition), 2010, 38(12): 79-83.
15	钟伟民, 李杰, 程辉, 等 . 基于FCM聚类的气化炉温度多模型软测量建模[J]. 化工学报, 2012, 63(12): 3951-3955.
	Zhong W M , Li J , Cheng H , et al . Multi-model soft-sensing modeling of gasifier temperature based on FCM clustering [J]. CIESC Journal, 2012, 63(12): 3951-3955.
16	Dhanachandra N , Manglem K , Chanu Y J . Image segmentation using K-means clustering algorithm and subtractive clustering algorithm[J]. Procedia Computer Science, 2015, 54: 764-771.
17	Gonzalez T F . Clustering to minimize the maximum intercluster distance [J]. Theoretical Computer Science, 1985, 38: 293-306.
18	Redmond S J , Heneghan C . A method for initialising the K-means clustering algorithm using kd-trees [J]. Pattern Recognition Letters, 2007, 28(8): 965-973.
19	Kapil S , Chawla M , Ansari M D . On K-means data clustering algorithm with genetic algorithm[C]//2016 Fourth International Conference on Parallel, Distributed and Grid Computing (PDGC). IEEE, 2016: 202-206.
20	Yu Q , Luo Y , Chen C , et al . Outlier-eliminated K-means clustering algorithm based on differential privacy preservation[J]. Applied Intelligence, 2016, 45(4): 1179-1191.
21	Deb A B , Dey L . Outlier detection and removal algorithm in K-means and hierarchical clustering[J]. World Journal of Computer Application and Technology, 2017, 5(2): 24-29.
22	Stumm W , Lee G F . Oxygenation of ferrous iron [J]. Industrial & Engineering Chemistry, 1961, 53(2): 143-146.
23	Jimenez C , Saez C , Martinez F , et al . Electrochemical dosing of iron and aluminum in continuous processes: a key step to explain electro-coagulation processes [J]. Separation & Purification Technology, 2012, 98(98): 102-108.
24	Lakshmanan D , Clifford D A , Samanta G . Ferrous and ferric ion generation during iron electrocoagulation [J]. Environmental Science & Technology, 2009, 43(10): 3853.
25	Sujana M G , Thakur R S , Rao S B . Removal of fluoride from aqueous solution by using alum sludge[J]. Journal of Colloid and Interface Science, 1998, 206(1): 94-101.
26	Zhou X , Yang C , Gui W . State transition algorithm [J]. Journal of Industrial and Management Optimization, 2012, 8(4): 1039-1056.
27	Gan G , Ng K P . k-means clustering with outlier removal[J]. Pattern Recognition Letters, 2017, 90: 8-14.
28	Ahmed M . A novel approach for outlier detection and clustering improvement[C]// Industrial Electronics & Applications. IEEE, 2013.
29	Zhou X , Gao D Y , Yang C . A comparative study of state transition algorithm with harmony search and artificial bee colony[C]//Proceedings of the Eighth International Conference on Bio-Inspired Computing: Theories and Applications (BIC-TA), 2013. Berlin, Heidelberg: Springer, 2013: 651-659.
30	刘靖明, 韩丽川, 侯立文 . 基于粒子群的K均值聚类算法[J]. 系统工程理论与实践, 2005, (6): 54-58.
	Liu J M , Han L C , Hou L W . K-means clustering algorithm based on particle swarm optimization [J]. Systems Engineering Theory and Practice, 2005, (6): 54-58.

算法	数据集	均值	方差	最优解
K均值	Iris	85.0108	354.9195	78.9408
	Wine	2.4714×10⁶	1.6264×10¹⁰	2.3707×10⁶
	CMC	2.3698×10⁴	50.0719	2.3690×10⁴
	Balance	3.4920×10³	627.7348	3.4723×10³
PSO-K均值	Iris	81.0750	131.6311	78.9408
	Wine	2.3758×10⁶	1.2698×10⁹	2.3707×10⁶
	CMC	2.3698×10⁴	32.7557	2.3690×10⁴
	Balance	3.4772×10³	27.6025	3.4723×10³
本文算法	Iris	78.9763	4.6151×10^-6	78.9521
	Wine	2.0401×10⁶	2.1555×10⁴	2.0400×10⁶
	CMC	2.3625×10⁴	0.7478	2.3624×10⁴
	Balance	3.4733×10³	6.9692	3.4723×10³

算法	数据集	均值	方差	最优解
K均值	Iris	85.0108	354.9195	78.9408
	Wine	2.4714×10⁶	1.6264×10¹⁰	2.3707×10⁶
	CMC	2.3698×10⁴	50.0719	2.3690×10⁴
	Balance	3.4920×10³	627.7348	3.4723×10³
PSO-K均值	Iris	81.0750	131.6311	78.9408
	Wine	2.3758×10⁶	1.2698×10⁹	2.3707×10⁶
	CMC	2.3698×10⁴	32.7557	2.3690×10⁴
	Balance	3.4772×10³	27.6025	3.4723×10³
本文算法	Iris	78.9763	4.6151×10^-6	78.9521
	Wine	2.0401×10⁶	2.1555×10⁴	2.0400×10⁶
	CMC	2.3625×10⁴	0.7478	2.3624×10⁴
	Balance	3.4733×10³	6.9692	3.4723×10³

方法	工况	q _max/ (mg/g)	K _L/ (L/mg)	k ₁	k ₂	k ₃
模型1		70.91	0.0951	5.9853×10⁶	1.04	1.73
模型2	1	73.51	0.0923	5.1347×10⁶	0.88	1.70
	2	70.32	0.1033	7.0325×10⁶	1.02	1.75
	3	68.89	0.0960	2.9871×10⁶	1.04	1.63
	4	69.67	0.0976	6.4153×10⁶	0.86	1.72
	5	70.52	0.0944	5.5360×10⁶	0.91	1.67
模型3	1	72.96	0.0962	4.1526×10⁶	0.98	1.66
	2	71.87	0.0965	5.7729×10⁶	0.95	1.68
	3	70.41	0.0974	7.2144×10⁶	0.87	1.73
	4	71.64	0.1036	6.9386×10⁶	1.05	1.75
	5	69.33	0.1012	6.1376×10⁶	0.83	1.71
模型4	1	72.62	0.0928	6.8358×10⁶	1.01	1.73
	2	69.48	0.0989	5.9037×10⁶	0.94	1.69
	3	70.59	0.0958	5.1075×10⁶	0.93	1.72
	4	71.03	0.1008	4.3165×10⁶	0.98	1.67
	5	69.75	0.0937	7.1049×10⁶	1.02	1.75

方法	工况	q _max/ (mg/g)	K _L/ (L/mg)	k ₁	k ₂	k ₃
模型1		70.91	0.0951	5.9853×10⁶	1.04	1.73
模型2	1	73.51	0.0923	5.1347×10⁶	0.88	1.70
	2	70.32	0.1033	7.0325×10⁶	1.02	1.75
	3	68.89	0.0960	2.9871×10⁶	1.04	1.63
	4	69.67	0.0976	6.4153×10⁶	0.86	1.72
	5	70.52	0.0944	5.5360×10⁶	0.91	1.67
模型3	1	72.96	0.0962	4.1526×10⁶	0.98	1.66
	2	71.87	0.0965	5.7729×10⁶	0.95	1.68
	3	70.41	0.0974	7.2144×10⁶	0.87	1.73
	4	71.64	0.1036	6.9386×10⁶	1.05	1.75
	5	69.33	0.1012	6.1376×10⁶	0.83	1.71
模型4	1	72.62	0.0928	6.8358×10⁶	1.01	1.73
	2	69.48	0.0989	5.9037×10⁶	0.94	1.69
	3	70.59	0.0958	5.1075×10⁶	0.93	1.72
	4	71.03	0.1008	4.3165×10⁶	0.98	1.67
	5	69.75	0.0937	7.1049×10⁶	1.02	1.75

方法	MRE	RMSE	MAXE/%
单一模型	0.1009	0.0047	29.0719
K均值-多模型	0.0848	0.0038	17.4880
PSO-K均值-多模型	0.0779	0.0037	17.3490
本文方法	0. 0561	0. 0027	13.9315