基于ANN的造纸废水处理智能控制系统研究

研究了人工神经网络技术在废纸造纸废水处理过程动态建模中应用的可行性,采用误差反向传播网络(BP网)建立了表征原水COD、加药量、进水流量、历史出水COD与预计出水COD之间复杂关系的动态模型和控制器模型,其计算输出值与过程的实际输出值具有较好的一致性。利用MCGS组态软件和所建模型搭建了一个造纸废水处理智能控制系统,对造纸厂现场排放废水进行实验,结果表明该系统对废水在线控制具有可行性。     

基于pso-SVM的废水厌氧处理过程软测量模型

由于厌氧消化过程的复杂性和厌氧菌的敏感性,保持厌氧消化体系的稳定和高效性是比较困难的.本文在实验室采用IC反应器构建了一套厌氧废水处理系统处理人工合成废水,基于支持向量机(SVM)提出了一种预测废水厌氧处理系统出水挥发性脂肪酸(VFA)浓度和COD去除率的软测量模型.为了提高模型的精确性和鲁棒性,加入pso算法(粒子群算法)优化SVM模型,并引入了分类策略对元数据集进行有效分类.仿真结果表明,基于pso-SVM模型的软测量模型对厌氧废水处理系统出水VFA浓度和COD去除率具有较好的预测能力,模型预测系统COD去除率及出水总VFA浓度测试样本数据相关系数分别为65.86%、85.25%;加入分类策略后,元数据集分成两类,模型预测系统COD去除率测试样本数据相关系数分别为92.34%、83.41%;模型预测系统出水总VFA浓度测试样本数据相关系数分别为99.14%、99.59%,系统预测精度明显提高.引入分类策略对元数据集进行有效分类,基于pso-SVM的软测量模型可为监控、优化和理解厌氧消化过程提供指导.     

造纸废水处理过程的高斯过程回归软测量建模

针对造纸废水处理系统的时变性、非线性和复杂性等特点,本文提出一种基于高斯过程回归的软测量模型.基于平方指数协方差、线性协方差和周期性协方差函数组合构建了7种高斯过程回归模型,分别对出水化学需氧量和出水悬浮固形物浓度进行回归预测.此外,还对比了多元线性回归模型、主成分回归模型、偏最小二乘模型、人工神经网络模型和高斯回归模型的预测效果.对比计算结果表明无论是对输出变量的训练拟合还是预测,高斯过程回归模型的拟合效果均优于非高斯过程回归模型.高斯过程回归模型的预测结果表明：对于出水化学需氧量,线性协方差函数与周期性协方差函数的组合模型可以取得最好的预测结果;对于出水悬浮固形物,平方指数协方差函数与线性协方差函数组合模型可以取得最好的预测结果.     

废纸造纸循环废水高效好氧菌的筛选和研究

利用平板分离方法,从废纸造纸封闭循环废水处理的生物接触氧化池中筛选出优势的高效菌株,采用形态学观察和生理生化试验发现主要由动胶菌属(Zoogloea sp.)和假单胞菌属(Pseudomona sp.)组成。用吸光光度法,确定其最佳生长时间。通过平行对比实验,在不同时间取样测定COD值,比较得出高效菌对废纸造纸循环废水的处理能力远高于普通菌。     

废纸造纸废水处理工程的设计与运行

文章针对废纸造纸生产中废水产生量大,SS、COD浓度较高的问题,将废纸造纸废水处理工程实践中的污水作为对象,研究一套废纸造纸废水回收利用的污水处理方法,减少再生纸造纸工业对环境的污染。实验结果和工程实践证明:利用混凝沉淀—活性污泥相结合工艺处理废纸造纸废水,达到一部分经物理处理后回收利用,一部分经生化处理后达标排放。采用该工艺处理效果好,SS去除率达99.7%、COD去除率达98.4%,且运行稳定,成本和一次性投资均较低,其排放废水水质满足《造纸工业水污染物排放标准》(GB3544-2001)中的排放要求。     

某饮料企业污水处理设施运行效果分析

对杭州某饮料企业废水处理设施运行效果进行了跟踪分析。分析表明对于该废水处理设施,其出水COD浓度和废水处理水量存在相关关系。当废水处理水量和进水COD浓度在一定范围时,设施可在最佳运行状态;处理量达到75%设计废水处理量下,应合理安排生产,控制进水COD值低于1 900 mg/L,才可保证出水水质达标排放。     

基于量子加权最小门限单元网络的出水COD预测

出水化学需氧量（COD）的快速准确测量对于污水处理过程水质的动态调控至关重要。针对出水COD难以实时检测的问题,提出一种基于量子加权最小门限单元（QWMGU）神经网络的出水COD预测方法。先通过多维单步（滑动窗口）预测技术构建时间序列;然后在最小门限单元（MGU）遗忘门、候选状态与输出环节设计量子计算机制,通过更新量子相移门矩阵替代MGU权值矩阵的更新,赋予网络神经元量子特性,并给出QWMGU模型设计的具体规则与构建流程。应用该方法对德州市污水处理厂2020年出水COD进行预测,并与5种流行预测模型进行对比,以检验模型优越性。结果表明：QWMGU网络的相对预测误差优于其他方法,且稳定性较高,其均方根误差、确定系数、平均绝对误差分别为0.073、1、0.047。该模型有助于实现污水处理厂COD的高效在线检测。

     

生化法处理造纸废水的试验研究

采用UASB反应器和A/O池组合的生化法进行了造纸废水处理的试验研究.结果表明,UASB反应器处理造纸废水,废水可生化性得到明显改善.COD去除率与水力停留时间呈正相关性,UASB反应器水力停留时间取24 h合适.该生化法COD去除率达93.33％,氨氮去除率达87.72％,系统出水COD浓度能满足造纸行业低档纸回用水要求.造纸废水中污染物成分比较复杂,经过UASB反应器和A/O池处理后有机成分变化比较明显,污染物种类梯级减少.通过A/O池处理,污染物减少一半,主要是小分子和低碳烷烃类,还含有少量难降解有机物.     

废纸造纸废水处理方法研究与工程应用

介绍了采用微滤+混凝沉淀处理废纸造纸废水的研究和工程实例。试验结果和工程实践证明,该工艺SS去除率达90%以上,COD去除率达85%以上,处理出水水质能达到国家排放标准(GB3544-2001),且具有设备简单、运行稳定、操作方便、成本低等特点。     

电解Fenton法处理生活垃圾焚烧厂渗滤液的动力学研究

采用三元电极,对垃圾焚烧厂渗滤液的生化-混凝出水进行了电解Fenton法处理COD反应动力学研究.结果表明:电解Fenton反应的COD降解曲线符合拟三级反应动力学,t时间的COD出水浓度可表示为:c_t=(2kt+c₀^-2)^-0.5.通过对不同工艺条件下电解Fenton法对渗滤液的COD降解曲线的统计分析,得出以电流密度、pH值、进水COD浓度为影响因子的k值经验模型.并将动力学模拟的COD出水值与实际运行的测量值进行比较,认为该模型可以较好地模拟电解Fenton法进行垃圾焚烧厂渗滤液后续处理的效果.     

Feasibility and simulation model of a pilot scale membrane bioreactor for wastewater treatment and reuse from Chinese traditional medicine

The lack and pollution of water resource make wastewater reuse necessary. The pilot scale long-term tests for submerged membrane bioreactor were conducted to treat the effluents of anaerobic or aerobic treatment process for the high-strength Chinese traditional medicine wastewater. This article was focused on the feasibility of the wastewater treatment and reuse at shorter hydraulic retention time （HRT） of 5.0, 3.2 and 2.13 h. MLSS growth, membrane flux, vacuum values and chemical cleaning periods were also investigated. The experimental results of treating two-phase anaerobic treatment effluent demonstrated that the CODfilt was less than 100 mg/L when the influent COD was between 500-10000 mg/L at HRT of 5.0 h, which could satisfy the normal discharged standard in China. The experimental results to treat cross flow aerobic reactor effluent demonstrated that the average value of CODfilt was 17.28 mg/L when the average value of influent COD was 192.84 mg/L at HRT of 2.13 h during 106 d, which could completely meet the normal standard for water reuse. The maximum MLSS and MLVSS reached 24000 and 14500 mg/L at HRT of 3.2 h respectively. Membrane flux had maximal resume degrees of 94.7% at vacuum value of 0.02 MPa after cleaning. Chemical cleaning periods of membrane module were 150 d. A simulation model of operational parameters was also established based on the theory of back propagation neural network and linear regression of traditional mathematical model. The simulation model showed that the optimum operational parameters were suggested as follows： HRT was 5.0 h, SRT was 100 d, the range of COD loading rate was between 10.664-20.451 kg/（m3.d）, the range of MLSS was between 7543-13694 mg/L.     

Short-term prediction of the influent quantity time series of wastewater treatment plant based on a chaos neural network model

By predicting influent quantity, a wastewater treatment plant (WWTP) can be well controlled. The nonlinear dynamic characteristic of WWTP influent quantity time series was analyzed, with the assumption that the series was predictable. Based on this, a short-term forecasting chaos neural network model of WWTP influent quantity was built by phase space reconstruction. Reasonable forecasting results were achieved using this method. Translated from Acta Scientiae Circumstantiae, 2006, 26(3): 416–419 [译自: 环境科学学报]     

城镇污水处理厂低浓度进水原因分析及提升措施

城镇污水处理厂进水浓度偏低是污水治理提质增效的难题之一。以无锡市某污水处理厂为例,基于水质分析和调研资料,采用水量平衡三角法分析了进水浓度偏低的原因分析,提出了整治措施。分析结果表明:该厂进水中地下水或河水渗入量和雨水混入量比例分别为28.9%和9.8%;晴天时进水浓度偏低的原因为污水管网渗漏和低浓度工业废水纳管,雨天时进水浓度偏低的主要原因是小区管网雨水混入,COD和BOD₅降幅分别可达到65.4%和52.9%。针对上述问题,提出了完善排水管理体制、工业企业排水实行三管分设整改、细化管网渗漏的检测排查和有效修复、加强管网验收和质量监管、尽量降低管网水位4项整治措施。     

填料气浮柱处理造纸废水的研究

采用填料气浮柱处理造纸行业纸机白水 ,通过对絮凝药剂碱式氯化铝的用量 ,废水进料表观流速、溶气水表观流速的大小 ,柱体回收段、分离段、浓缩段的深度和废水进料位置等因素的考察 ,表明在一定的条件下 ,该设备和工艺对降低废水CODCr和悬浮物含量 ,提高浮渣污泥固含量等指标具有明显的效果。同时 ,连续试验结果也说明该设备对降低废水BOD5 也有较好的作用。     

四川某中小城镇污水处理厂进水水质统计分析

生活污水水质对处理工艺的选择、设计参数的选取以及稳定运行十分重要。运用SPSS软件,对四川某中小城镇污水处理厂为期一年的进水指标(水量、COD、氨氮、总磷、pH、SS)进行方差和相关性分析。方差分析结果显示进水量、氨氮和SS在月份和季节之间存在显著性差异,而总磷和pH仅在月份之间存在显著差异,COD在季度和月份之间无显著性差异;指标相关性分析表明,进水量和氨氮之间呈负相关,氨氮和总磷、COD和总磷之间呈正相关,且相关性显著。在此分析的基础上对污水处理厂设计、工艺选择及运行管理提出了一些建议。     

基于模糊神经网络的A2/O工艺出水氨氮在线预测模型

采用厌氧/缺氧/好氧污水处理系统(A2/O)对人工合成污水进行处理,并利用人工神经网络(ANN)模型和自适应模糊人工神经网络(ANFIS)模型对A2/O处理污水的过程进行仿真模拟.在MATLAB环境下,选取可在线监测的水力停留时间(HRT)、进水pH值(pH)、好氧池溶解氧(DO)和混合液回流比(r)作为输入参量,系统出水氨氮浓度(NH4+eff)为输出量,建立在线预测模型.结合自适应模糊C均值聚类算法,确定ANFIS模型的模糊规则数及最优运行参数,对实验数据进行仿真预测.结果表明,与ANN模型相比,ANFIS模型的仿真输出值与实际值的拟合程度更高,相对误差在6.45%之内,平均绝对百分比误差(MAPE)为2.8%,均方根误差(RMSE)为0.1209,相关系数(R)达0.9956.模型训练过程中所得到的三维曲面图,可直观的反映各因素与出水氨氮浓度之间的非线性函数关系,为A2/O系统的高效稳定运行提供指导.     

两段活性污泥法处理味精废水的中试研究

采用两段活性污泥法处理味精废水 ,试验规模 0 5m3/h ,当进水CODCr浓度 16 80～ 70 30mg/L,NH3 N浓度 2 0 6～ 1999mg/L时 ,经石灰中和 ,空气吹脱对离交废水的预处理后 ,出水CODCr<10 0 0mg/L。试验中 ,兼氧池和一段好氧池污泥浓度保持在 6 0 0 0～ 80 0 0mg/L之间 ,二段好氧池保持在 40 0 0～ 6 0 0 0mg/L之间。生物处理总停留时间 5 0h。在离交废水预处理中 ,pH中和至 9 5～ 10 ,鼓气量在 10 0m3/h左右 ,水温加至 5 5℃左右 ,经 8h ,可将原水NH3 N从12 0 0 0mg/L左右 ,脱除 6 5 %以上 ,出水氨氮可达 40 0 0mg/L左右。     

Simulation and control strategy for the variational influent of WWTP

With the development of activated sludge model, the simulation software for the design and operation of wastewater treatment plant （WWTP） was produced and has been widely used. The dynamic change of the quality and flow of influent are major factors causing the unstable operation of wastewater treatment process. As a basic model, ASMI model was used for the simulation of activated sludge process, and double exponential model was selected for the simulation of secondary sedimentation tank. The influences of influent change to the aeration tank and secondary sedimentation tank were investigated, and the relationship among influent change, the quality of effluent and the level of sludge blanket in secondary sedimentation tank was established. On the basis of the simulation results, the operation of the WWTP could be adjusted under the dynamic change of the influent. Furthermore, the controlling strategy combined the feed-forward on the influent flow and the feedback on the level of sludge blanket in the secondary sedimentation tank was studied.     

Pollutants removal and simulation model of combined membrane process for wastewater treatment and reuse in submarine cabin for long voyage

A laboratory scale test was conducted in a combined membrane process (CMP) with a capacity of 2.91 m3/d for 240 d to treat the mixed wastewater of humidity condensate, hygiene wastewater and urine in submarine cabin during prolonged voyage. Removal performance of chemical oxygen demand (COD), ammonia nitrogen (NH4 +-N), turbidity and anionic surfactants (LAS) was investigated under di erent conditions. It was observed that the e uent COD, NH4 +-N, turbidity and LAS flocculated in ranges of 0.19–0.85 mg/L, 0.03–0.18 mg/L, 0.0–0.15 NTU and 0.0–0.05 mg/L, respectively in spite of considerable fluctuation in corresponding influent of 2120–5350 mg/L, 79.5–129.3 mg/L, 110–181.1NTU and 4.9–5.4 mg/L. The e uent quality of the CMP could meet the requirements of mechanical water and hygiene water according to the class I water quality standards in China (GB3838-2002). The removal rates of COD, NH4 +-N, turbidity and LAS removed in the MBR were more than 90%, which indicated that biodegradation is indispensable and plays a major role in the wastewater treatment and reuse. A model, built on the back propagation neural network (BPNN) theory, was developed for the simulation of CMP and produced high reliability. The average error of COD and NH4 +-N was 5.14% and 6.20%, respectively, and the root mean squared error of turbidity and LAS was 2.76% and 1.41%, respectively. The results indicated that the model well fitted the laboratory data, and was able to simulate the removal of COD, NH4 +-N, turbidity and LAS. It also suggested that the model proposed could reflect and manage the operation of CMP for the treatment of the mixed wastewaters in submarine.