基于二次分解和BiLSTM的污水厂出水COD浓度预测

doi:10.11949/0438-1157.20241285

摘要/Abstract

摘要：

针对某污水处理厂出水化学需氧量（chemical oxygen demand，COD）浓度预测问题，提出了一种基于自适应噪声完备经验模态分解（complete ensemble empirical mode decomposition with adaptive noise，CEEMDAN）、变分模态分解（variational mode decomposition，VMD）二次分解、双向长短期记忆神经网络（bidirectional long short-term memory，BiLSTM）的出水COD预测模型，并引入吸血水蛭优化算法（blood-sucking leech optimizer，BSLO）对模型进行优化。首先，设计CEEMDAN算法对原始出水浓度序列进行分解，将复杂的时间序列分解为若干相对简单的子序列；然后，应用VMD对具有不稳定的高频不规则波形的子序列进行二次分解；最后，应用BSLO对BiLSTM进行优化，并比较未分解、一次分解、二次分解以及无优化算法下80个模型在污水厂出水COD浓度预测问题中的性能。结果表明，优化算法的引入提高了模型预测的性能，BSLO模型具有更快的速度和更高的精度；相比于其他模型，基于二次分解的BSLO+CEEMDAN+VMD+BiLSTM模型能够有效克服实测数据非线性和复杂性问题，在该厂出水COD预测中表现出优秀的预测精度和泛化能力。

关键词: 算法, 预测, 废水, 优化, 污水处理

Abstract:

For the prediction of chemical oxygen demand (COD) concentration in the effluent of a sewage treatment plant, a COD prediction model based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and variational mode decomposition (VMD) secondary decomposition and bidirectional long short-term memory (BiLSTM) neural network is proposed, and the blood-sucking optimization algorithm (blood-sucking leech optimizer, BSLO) is used to optimize the model. Firstly, design the CEEMDAN algorithm to decompose the original effluent concentration sequence, breaking down the complex time series into several relatively simple sub sequences. Then, VMD is applied to perform quadratic decomposition on subsequences with unstable high-frequency irregular waveforms. Finally, BSLO was applied to optimize BiLSTM, and the performance of 80 models in predicting COD concentration in wastewater treatment plant effluent was compared under undecomposed, first decomposed, second decomposed, and non optimized algorithms. The results show that the introduction of optimization algorithms improves the performance of model prediction, and the BSLO model has faster speed and higher accuracy. Compared with other models, the BSLO+CEEMDAN+VMD+BiLSTM model based on quadratic decomposition can effectively overcome the nonlinearity and complexity of measured data, and exhibits excellent prediction accuracy and generalization ability in the COD prediction of the effluent from the plant.

Key words: algorithm, prediction, waste water, optimization, wastewater treatment

中图分类号:

TP 183

张京新, 何皎洁, 蔡庆旺, 康子怡, 杨玉思, 王彤, 曹仙桃, 杨利伟. 基于二次分解和BiLSTM的污水厂出水COD浓度预测[J]. 化工学报, 2025, 76(6): 2859-2871.

Jingxin ZHANG, Jiaojie HE, Qingwang CAI, Ziyi KANG, Yusi YANG, Tong WANG, Xiantao CAO, Liwei YANG. Prediction of COD concentration in wastewater treatment plant effluent based on secondary decomposition and BiLSTM[J]. CIESC Journal, 2025, 76(6): 2859-2871.

图/表 12

图1 LSTM结构图

Fig.1 Structure of the LSTM

图2 BiLSTM结构图

Fig.2 Structure of the BiLSTM

图3 基于二次分解的BiLSTM模型

Fig.3 BiLSTM model based on quadratic decomposition

图4 出水COD浓度日监测数据

Fig.4 Daily monitoring data of effluent COD concentration

图5 出水COD浓度自相关性

Fig.5 Autocorrelation of effluent COD concentration

图6 CEEMDAN分解得到的子序列

Fig.6 Subsequences obtained from CEEMDAN decomposition

图7 最大中心频率变化曲线

Fig.7 Maximum center frequency variation curve

图8 不同优化算法的适应度曲线

Fig.8 Fitness curves of different optimization algorithms

图9 VMD分解得到的子序列

Fig.9 Subsequences obtained from VMD decomposition

图10 预测模型效果

Fig.10 Prediction model performance

表1 CEEMDAN分解子序列模型的决定系数

Table 1 The model determination coefficient of CEEMDAN decomposition subsequence

原始数据分解的子序列	BSLO+LSTM			BSLO+BiLSTM
原始数据分解的子序列	训练集	验证集	测试集	训练集	验证集	测试集
IMF₁	0.4809	0.1748	0.4141	0.4667	0.1650	0.3964
IMF₂	0.8590	0.7568	0.7047	0.8571	0.7564	0.7035
IMF₃	0.8845	0.8497	0.8275	0.8842	0.8432	0.8241
IMF₄	0.9791	0.9809	0.9839	0.9783	0.9814	0.9830
IMF₅	0.9943	0.9940	0.9933	0.9944	0.9942	0.9941
IMF₆	0.9975	0.9991	0.9987	0.9981	0.9993	0.9986
IMF₇	0.9976	0.9989	0.9982	0.9981	0.9989	0.9983
IMF₈	0.9989	0.9996	0.9994	0.9984	0.9996	0.9995
IMF₉	0.9989	0.9999	0.9998	0.9990	0.9998	0.9997
IMF₁₀	0.9987	0.9332	0.8302	0.9993	0.9459	0.8734
RES	0.9999	0.9267	0.6286	0.9999	0.9748	0.8686

表2 二次分解子序列模型的决定系数

Table 2 The coefficient of decomposition of the secondary decomposition subsequence model

二次分解的子序列	BSLO+LSTM			BSLO+BiLSTM
二次分解的子序列	训练集	验证集	测试集	训练集	验证集	测试集
imf₁	0.9736	0.9777	0.9736	0.9945	0.9954	0.9950
imf₂	0.9732	0.9791	0.9583	0.9763	0.9823	0.9620
imf₃	0.9735	0.9826	0.9783	0.9761	0.9849	0.9785
imf₄	0.9751	0.9762	0.9640	0.9746	0.9753	0.9621
imf₅	0.9638	0.9781	0.9721	0.9668	0.9796	0.9754
imf₆	0.9357	0.9225	0.9383	0.9457	0.9334	0.9474
res	0.2731	0.3081	0.0513	0.2576	0.2953	0.0423

参考文献 47

[1]	Tayyab M, Jemai J, Lim H, et al. A sustainable development framework for a cleaner multi-item multi-stage textile production system with a process improvement initiative[J]. Journal of Cleaner Production, 2020, 246: 119055.
[2]	汪锐, 余雅丹, 潘志成, 等. 模拟预测模型在污水处理中的应用: 现状与挑战[J]. 水处理技术, 2022, 48(6): 20-23, 29.
	Wang R, Yu Y D, Pan Z C, et al. Application of simulated predictive models to wastewater treatment: status and challenges[J]. Technology of Water Treatment, 2022, 48(6): 20-23, 29.
[3]	柴伟, 郭龙航, 池彬彬. 污水处理厂出水水质变量区间预测建模[J]. 化工学报, 2019, 70(9): 3449-3457.
	Chai W, Guo L H, Chi B B. Interval model for predicting effluent quality variables of wastewater treatment plants[J]. CIESC Journal, 2019, 70(9): 3449-3457.
[4]	郭鑫, 李文静, 乔俊飞. 基于自组织模块化神经网络的污水处理过程出水参数预测[J]. 化工学报, 2024, 75(9): 3242-3254.
	Guo X, Li W J, Qiao J F. Prediction of effluent parameters in wastewater treatment process using self-organizing modular neural network[J]. CIESC Journal, 2024, 75(9): 3242-3254.
[5]	张勇, 赵景波, 权利敏. 基于卷积层-注意力机制的长短期记忆网络出水氨氮浓度预测方法[J]. 化工学报, 2024, 75(12): 4679-4688.
	Zhang Y, Zhao J B, Quan L M. Prediction method of ammonia nitrogen concentration in effluent based on convolution layer-attention mechanism in long-term and short-term memory networks[J]. CIESC Journal, 2024, 75(12): 4679-4688.
[6]	梅丹, 张恒. 基于VMD和深度学习的污水出水水质预测[J/OL]. 长江科学院院报, 2024. .
	Mei D, Zhang H. Prediction of wastewater quality based on VMD and deep learning[J/OL]. Journal of Yangtze River Scientific Research Institute, 2024. .
[7]	李洪澎, 周平. 基于两阶段邻域搜索的污水处理过程智能优化控制[J]. 控制理论与应用, 2024, 41(9): 1569-1577.
	Li H P, Zhou P. Intelligent optimization control of wastewater treatment process based on two-stage neighborhood search[J]. Control Theory & Applications, 2024, 41(9): 1569-1577.
[8]	史红伟, 陈祺, 王云龙, 等. 基于encoder-decoder框架的城镇污水厂出水水质预测[J]. 中国农村水利水电, 2023(11): 93-99.
	Shi H W, Chen Q, Wang Y L, et al. Water quality prediction of town sewage plant effluent based on encoder-decoder framework[J]. China Rural Water and Hydropower, 2023(11): 93-99.
[9]	安昱宁, 朱四富, 刘静, 等. 基于深度学习的污水处理厂出水总磷预测方法[J]. 工业水处理, 2024, 44(10): 143-150.
	An Y N, Zhu S F, Liu J, et al. Prediction of total phosphorus in wastewater treatment plant effluent based on deep learning[J]. Industrial Water Treatment, 2024, 44(10): 143-150.
[10]	姚怡帆, 荆玉姝, 王丽艳, 等. 基于集成模型的污水处理厂出水总氮预测方法[J]. 工业水处理, 2023, 43(9): 187-194.
	Yao Y F, Jing Y S, Wang L Y, et al. Prediction method of total nitrogen in wastewater treatment plant based on integrated model[J]. Industrial Water Treatment, 2023, 43(9): 187-194.
[11]	Nourani V, Asghari P, Sharghi E. Artificial intelligence based ensemble modeling of wastewater treatment plant using jittered data[J]. Journal of Cleaner Production, 2021, 291: 125772.
[12]	张净, 窦慧芸, 蒋武, 等. 基于PIWT-IPSO-BP的污水厂出水COD含量的预测模型[J]. 中国农村水利水电, 2024(9): 15-20, 28.
	Zhang J, Dou H Y, Jiang W, et al. Predictive model of COD content in wastewater plant effluent based on PIWT-IPSO-BP[J]. China Rural Water and Hydropower, 2024(9): 15-20, 28.
[13]	陈霖, 刘浩威, 王庆宏, 等. 基于机器学习算法的炼化污水厂出水水质预测模型研究[J]. 工业水处理, 2024. DOI: 10.19965/j.cnki.iwt.2024-0515 .
	Chen L, Liu H W, Wang Q H, et al. Study on prediction model of effluent quality of refinery wastewater plant based on machine learning algorithm[J]. China Industrial Economics, 2024. DOI: 10.19965/j.cnki.iwt.2024-0515 .
[14]	乔俊飞, 安茹, 韩红桂. 基于RBF神经网络的出水氨氮预测研究[J]. 控制工程, 2016, 23(9): 1301-1305.
	Qiao J F, An R, Han H G. Water ammonia nitrogen prediction research based on RBF neural network[J]. Control Engineering of China, 2016, 23(9): 1301-1305.
[15]	Xie Y F, Chen Y Q, Wei Q, et al. A hybrid deep learning approach to improve real-time effluent quality prediction in wastewater treatment plant[J]. Water Research, 2024, 250: 121092.
[16]	Meng X, Zhang Y, Qiao J F. An adaptive task-oriented RBF network for key water quality parameters prediction in wastewater treatment process[J]. Neural Computing and Applications, 2021, 33(17): 11401-11414.
[17]	Wongburi P, Park J K. Prediction of wastewater treatment plant effluent water quality using recurrent neural network (RNN) models[J]. Water, 2023, 15(19): 3325.
[18]	Zhu J J, Kang L L, Anderson P R. Predicting influent biochemical oxygen demand: balancing energy demand and risk management[J]. Water Research, 2018, 128: 304-313.
[19]	Yu Y D, Chen Y W, Huang S B, et al. Enhancing the effluent prediction accuracy with insufficient data based on transfer learning and LSTM algorithm in WWTPs[J]. Journal of Water Process Engineering, 2024, 62: 105267.
[20]	貟姗, 王磊, 万立国, 等. 化学需氧量电化学传感器研究[J]. 分析化学, 2024, 52(9): 1298-1306.
	Yun S, Wang L, Wan L G, et al. Research on electrochemical chemical oxygen demand sensor[J]. Chinese Journal of Analytical Chemistry, 2024, 52(9): 1298-1306.
[21]	刘浩威, 陈霖, 李巨峰, 等. 基于数据驱动的污水处理系统水质预测研究进展[J]. 工业水处理, 2025, 45(4): 1-9.
	Liu H W, Chen L, Li J F, et al. Research progress of water quality prediction of sewage treatment system based on data driving[J]. China Industrial Economics, 2025, 45(4): 1-9.
[22]	薛桂香, 苗敬礼, 张丹, 等. 基于VMD-CNN-BiLSTM人工智能算法的节段拼装自复位桥梁地震响应预测方法[J/OL]. 中国公路学报, 2024. .
	Xue G X, Miao J L, Zhang D, et al. Prediction method of seismic response of segmental self-resetting bridge based on VMD-CNN-BiLSTM artificial intelligence algorithm[J/OL]. China Journal of Highway and Transport, 2024. .
[23]	王雷, 张煜, 赵艺琨, 等. 基于多空间维度联合方法改进的BiLSTM出水氨氮预测方法[J]. 中国农村水利水电, 2024(2): 17-24.
	Wang L, Zhang Y, Zhao Y K, et al. Improved BiLSTM effluent ammonia nitrogen prediction method based on multi-dimensional joint method[J]. China Rural Water and Hydropower, 2024(2): 17-24.
[24]	荣统瑞, 侯恩科, 夏冰冰. 基于二次分解和BO-BiLSTM组合模型的采煤工作面瓦斯涌出量预测方法研究[J]. 煤矿安全, 2024, 55(5): 83-92.
	Rong T R, Hou E K, Xia B B. Research on prediction method of coal mining face gas outflow based on quadratic decomposition and BO-BiLSTM combination model[J]. Safety in Coal Mines, 2024, 55(5): 83-92.
[25]	施艳艳, 杨珍, 王萌, 等. 基于多特征混合与GWO-SVM的气液两相流流型识别方法[J]. 化工学报, 2024, 75(10): 3536-3547.
	Shi Y Y, Yang Z, Wang M, et al. Flow pattern identification method based on multi-feature extraction and GWO-SVM for gas-liquid two-phase flow[J]. CIESC Journal, 2024, 75(10): 3536-3547.
[26]	张弛, 李浩, 胡海涛, 等. 基于自动特征工程的飞行器轴承故障诊断[J]. 化工学报, 2021, 72(S1): 430-436, 569.
	Zhang C, Li H, Hu H T, et al. Aircraft bearing fault diagnosis based on automatic feature engineering[J]. CIESC Journal, 2021, 72(S1): 430-436, 569.
[27]	杨新, 麻哲瑞, 申赫男, 等. 基于多尺度特征能量-核极限学习机的双循环流化床气流堵塞故障智能诊断[J]. 化工学报, 2019, 70(7): 2616-2625.
	Yang X, Ma Z R, Shen H N, et al. Fault diagnosis of airflow jamming fault in double circulating fluidized bed based on multi-scale feature energy and KELM[J]. CIESC Journal, 2019, 70(7): 2616-2625.
[28]	胡锐, 乔加飞, 李永华, 等. 基于WOA-VMD-SSA-LSTM的中长期风电预测[J]. 太阳能学报, 2024, 45(9): 549-556.
	Hu R, Qiao J F, Li Y H, et al. Medium and long term wind power forecast based on WOA-VMD-SSA-LSTM[J]. Acta Energiae Solaris Sinica, 2024, 45(9): 549-556.
[29]	盛四清, 袁伟智. 基于二次分解和BiLSTM-XGBoost-ARIMA组合模型的短期风速预测[J/OL]. 华北电力大学学报(自然科学版), 2024. .
	Sheng S Q, Yuan W Z. Short-term wind speed prediction based on quadratic decomposition and BiLSTM-XGBoost-ARIMA combined model[J/OL]. Journal of North China Electric Power University (Natural Science Edition), 2024. .
[30]	杨帆, 江松, 黄光球, 等. 基于数据分解和权重优化的VOCs浓度区间预测[J]. 环境科学学报, 2024, 44(11): 50-61.
	Yang F, Jiang S, Huang G Q, et al. VOCs concentration interval prediction based on data decomposition and weight optimization[J]. Acta Scientiae Circumstantiae, 2024, 44(11): 50-61.
[31]	李文华, 叶洪涛, 罗文广, 等. 基于MHSA-LSTM的软测量建模及其在化工过程中的应用[J]. 化工学报, 2024, 75(12): 4654-4665.
	Li W H, Ye H T, Luo W G, et al. Soft sensor modeling based on MHSA-LSTM and its application in chemical process[J]. CIESC Journal, 2024, 75(12): 4654-4665.
[32]	马会彪, 曾静. 基于POD-LSTM的污水处理过程模型预测控制[J]. 电子测量技术, 2024, 47(13): 81-88.
	Ma H B, Zeng J. Model predictive control of sewage treatment process based on POD-LSTM[J]. Electronic Measurement Technology, 2024, 47(13): 81-88.
[33]	刘立邦, 杨颂, 王志坚, 等. 基于改进WOA-LSTM的焦炭质量预测[J]. 化工学报, 2022, 73(3): 1291-1299.
	Liu L B, Yang S, Wang Z J, et al. Prediction of coke quality based on improved WOA-LSTM[J]. CIESC Journal, 2022, 73(3): 1291-1299.
[34]	邓明亮, 张钊, 周红艳, 等. 基于PCA-PSO_KFCM聚类和BiLSTM-Attention的短期电力负荷预测[J/OL]. 计算机工程与科学, 2024. .
	Deng M L, Zhang Z, Zhou H Y, et al. Short-term power load forecasting based on PCA-PSO _ KFCM clustering and BiLSTM-Attention[J/OL]. Computer Engineering & Science, 2024. .
[35]	王地欣, 王佳昊, 李敏, 等. 面向水声通信网络的异常攻击检测[J]. 计算机应用, 2025, 45(2): 526-533.
	Wang D X, Wang J H, Li M, et al. Abnormal attack detection for underwater acoustic communication network[J]. Journal of Computer Applications, 2025, 45(2): 526-533.
[36]	闫琳琦, 王振雷. 基于STA-BiLSTM-LightGBM组合模型的多步预测软测量建模[J]. 化工学报, 2023, 74(8): 3407-3418.
	Yan L Q, Wang Z L. Multi-step predictive soft sensor modeling based on STA-BiLSTM-LightGBM combined model[J]. CIESC Journal, 2023, 74(8): 3407-3418.
[37]	Bai J F, Nguyen-Xuan H, Atroshchenko E, et al. Blood-sucking leech optimizer[J]. Advances in Engineering Software, 2024, 195: 103696.
[38]	刘倩, 王兆茹, 胥寒莉, 等. 基于粒子群优化XGBoost模型的PM_2.5浓度反演[J]. 环境科学, 2024. DOI: 10.13227/j.hjkx.202407279 .
	Liu Q, Wang Z R, Xu H L, et al. Inversion of PM_2.5 concentration based on particle swarm optimization XGBoost model[J]. China Industrial Economics, 2024. DOI: 10.13227/j.hjkx.202407279 .
[39]	Xue J K, Shen B. Dung beetle optimizer: a new meta-heuristic algorithm for global optimization[J]. The Journal of Supercomputing, 2023, 79(7): 7305-7336.
[40]	Dehghani M, Hubálovský Š, Trojovský P. Northern goshawk optimization: a new swarm-based algorithm for solving optimization problems[J]. IEEE Access, 2021, 9: 162059-162080.
[41]	Su H, Zhao D, Heidari A A, et al. RIME: a physics-based optimization[J]. Neurocomputing, 2023, 532: 183-214.
[42]	孙伟峰, 李雯, 范陈清, 等. 基于AR模型的多普勒散射计回波功率谱估计[J]. 遥感信息, 2023, 38(2): 40-47.
	Sun W F, Li W, Fan C Q, et al. An echo power spectral density estimation method based on AR model for Doppler scatterometter[J]. Remote Sensing Information, 2023, 38(2): 40-47.
[43]	孟一恒, 黄永明, 卢永, 等. 基于ARMA模型磁暴抑制的地震TEC异常分析[J]. 防灾减灾工程学报, 2023, 43(4): 896-904.
	Meng Y H, Huang Y M, Lu Y, et al. Analysis of anomalies in seismic TEC data based on ARMA magnetic storm suppression model[J]. Journal of Disaster Prevention and Mitigation Engineering, 2023, 43(4): 896-904.
[44]	郑庆锋, 梁萍, 段玉森, 等. 2006—2021年夏半年上海臭氧浓度特征及其大气环流背景分析[J]. 环境科学, 2024, 45(2): 645-654.
	Zheng Q F, Liang P, Duan Y S, et al. Characteristics of ozone concentration in Shanghai and its associated atmospheric circulation background during summer half-years from 2006 to 2021[J]. Environmental Science, 2024, 45(2): 645-654.
[45]	刘丽娜, 秦飞跃, 宋欣钢. 基于状态空间模型的加工中心功能铣头实时寿命预测[J]. 机床与液压, 2023, 51(24): 184-187.
	Liu L N, Qin F Y, Song X G. Real-time life prediction of functional milling head of machining center based on state-space model[J]. Machine Tool & Hydraulics, 2023, 51(24): 184-187.
[46]	严慧峰, 黄定疆, 谢垚, 等. 短期电力负荷预测模型的比较研究[J]. 中国科学技术大学学报, 2019, 49(2): 119-124.
	Yan H F, Huang D J, Xie Y, et al. Comparative study of short-term electrical load forecast models[J]. Journal of University of Science and Technology of China, 2019, 49(2): 119-124.
[47]	王龙洋, 蒙西, 乔俊飞. 基于改进集合经验模态分解和深度信念网络的出水总磷预测[J]. 化工学报, 2021, 72(5): 2745-2753.
	Wang L Y, Meng X, Qiao J F. Prediction of effluent total phosphorus based on modified ensemble empirical mode decomposition and deep belief network[J]. CIESC Journal, 2021, 72(5): 2745-2753.