导　　师： 罗飞

学科专业： 081101

授予学位： 博士

作　　者： ;

机构地区： 华南理工大学

摘　　要： 污水生化处理是当前和未来很长一段时间污水处理和水污染治理应用最广泛、最有效和最经济的技术。城市污水处理过程是一个复杂的生物与化学过程，具有多变量、随机性、非线性、模糊性、时变性、多目标等特性，传统的控制方法不能有效地解决这些问题，以至难于进行系统优化，因此研发污水智能控制系统是解决当前污水处理厂优化控制问题的关键。本课题就是解决污水生化处理智能控制系统运行控制的若干个难题。 本课题来源于国家自然科学基金项目“污水生化处理系统的建模与节能优化控制”/(项目编号：60774032/)、国家住房城乡建设部“污水处理厂污泥回流系统的建模与节能优化控制系统”/(项目编号：2010-K9-47/)、广州市科技与信息化局科技支撑项目“基于智能控制技术的节能型污泥回流系统的开发与产业化”/(项目编号：2010Z1-E301/)、广州市教育局“节能减排/(水处理/)自动化技术应用研究创新学术团队”（穗教科/[2009/]11号）。 本课题针对污水生化处理智能控制的出水水质预测、曝气控制、污泥浓度控制和污水处理运行的节能优化控制四个方面开展研究，分别使用拉普拉斯特征映射与支持向量机相结合（Laplacian Eigenmap-Support Vector Machine，LE-SVM）的算法进行软测量建模对出水水质进行预测；对曝气系统进行一种全局渐近稳定的自适应神经网络控制；对污泥浓度实行鲁棒直接自适应模糊控制；以出水指标作为边界条件，对曝气及污泥回流、排放的费用最优化采用模糊离散粒子群算法进行动态优化，以达到整个污水处理厂的运行费用最低的优化控制，通过仿真验证取得了较好地效果，为污水厂的节能优化控制提供了理论依据，取得了一定的研究成果，具有理论与现实意义。 本文的主要研究内容概括如下： 1.针对污水处理工艺参数众多，包括水质、水量、负荷、污泥性能等数十个指标相互关联等问题，本文采用拉普拉斯特征映射与支持向量机的相结合的方法建立软测量模型，对出水水质参数生物化学需氧量（Biochemical Oxygen Demand，BOD）进行预测。首先使用拉普拉斯特征映射对测量数据进行非线性降维，解决数据相互之间的相关问题，然后应用支持向量机建模方法对水质参数进行建模预测。利用污水厂采集数据仿真结果表明，该方法建立污水处理软测量模型相比RBF（Radial Base Function）神经网络有更好的预测精度和泛化能力。 2.针对于曝气系统中溶解氧的传递系数这个控制参数增益的不确定性，提出一种全局渐近稳定的自适应神经控制（adaptive neural control, ANC）策略，使用线性参数化神经网络逼近闭环系统的总体不确定项。首先提出一种可变控制增益的比例微分（proportional differential, PD）控制方法用于全局镇定被控对象。然后利用状态变化解决由未知控制增益函数导致的控制奇异问题。最后设计了一种连续的自适应鲁棒控制项以实现闭环系统的渐近跟踪。与现有的全局或渐近跟踪ANC方法相比较，本文所提方法不仅简化了PD控制增益的选择，而且减轻了控制输入的颤振问题。仿真实验结果表明了本文所提方法相比传统的PI（proportional integral）控制方法能够有效地动态稳定溶解氧的浓度。 3.针对污水处理系统污浓度的控制问题中污泥回流、排泥系统中控制方向的不确定性，给出一种新的鲁棒直接自适应模糊控制算法。首先，采用基于Lyapunov的理想控制律新形式解决了控制奇异问题。其次，利用Nussbaum函数以及新的推导方式解决了非线性系统的未知控制方向问题。此外，使用e2‐修正代替之前的σ‐修正设计模糊系统的参数自适应律，这不仅获得了自适应参数的有界性，而且实现了跟踪误差的渐近收敛到零。最后，仿真结果表明此方法能够通过回流污泥、排放污泥有效地稳定生化池的污泥浓度，验证了本文所提方法的有效性。 4.针对污水处理过程运行费用最优控制问题，以溶解氧浓度和污泥回流量作为控制变量，以曝气的电耗、回流泵的电耗和剩余污泥处置费用之和作为目标函数，以出水水质和有机物排放总量作为约束条件，提出了模糊离散粒子群（fuzzy discrete particle swarm，FDPSO）算法进行优化控制，寻找污水处理费用的最优值。在同一约束前提下，用本文所提算法与粒子群算法、遗传算法寻优过程相比较。结果表明，该方法能有效提高粒子群的多样性，具有可靠的全局收敛性及较快的收敛速度，搜索到的费用最优值的次数最多，运行费用平均值和方差最小，能在污水处理过程的动态环境下完成有效的寻优，适合应用于污水处理过程的最优控制中。 Biological wastewater treatment process isone of the most widely，the most effective andeconomic technology in the field of wastewater treatment and water pollution control incurrent and future for a long time。For urban wastewater treatment process is a complexbiological and chemical process，multivariable and randomness ambiguity, time-varying,nonlinear and multiple objective features,the application of traditional control method cannoteffectively solve these problems, so it is hard to optimize system. Therefore, the intelligentcontrol system which takes optimal water quality and energy conservation as the goal is thekey to solve the problem of optimization control in the in the wastewater treatment plant atpresent.This topic is to solve the wastewater biochemical treatment system operation controlof a few problems is based on the theory of intelligent control technique。 This thesis comes from National Natural Science Foundation of China /(No.60774032/)，Special Research Fund of Ministry of Housing and Urban-Rural Development of China underGrant /(No.2010-k9-47/)，and Key Project of Guangzhou Scientific Program of China underGrant （No.2010Z1-E301），‘Fund of Innovation Creation Academy Group’ established by theGuangzhou Education Bureau（No.2009-11）。 This topic in the process of bio chemical wastewater treatment for the control of the fouraspects of research: the water quality control，aeration control， sludge concentration controland energy-saving optimization control, Respectively using intelligent algorithm for softmeasurement modeling to predict water quality control; For aeration system is a kind ofglobal asymptotic stability of the adaptive neural network control; On the sludgeconcentration implement robust direct adaptive fuzzy control; Finally water quality asrestriction factorsas the boundary conditions of the ost of aeration and the recyle sludge,emission optimization fuzzy discrete particle swarm optimization /(fdpso/) algorithm fordynamic optimization, in order to achieve the whole wastewater treatment plant of the lowestoperating cost optimization control, achieved better effect is validated by computer simulation,for wastewater plant energy saving optimization control provides theory basis, has obtainedcertain research results, has theoretical and practical significance。 The main contents of the thesis are outlined as follows. 1. Many wastewater treatment process parameters, including water quality, water quantity,load, sludge properties, such as dozens of indicators relate to each other,in this paper, the wastewater plantbiochemical oxygen demand predictive control based on laplacian eigenmap/(LE/) and support vector machine /(SVM/) method to establish the soft measurement model。First use of Laplace feature mapping was carried out on the measurement data processing,solve the problem of data relative between each other, so as to improve the accuracy androbustness of the model. At the same time, the application of support vector machine /(SVM/)modeling method to improve the generalization ability of water quality soft measurementmodel. Using wastewater collection data, the simulation results show that wastewatertreatment soft measurement model based on support vector machine /(SVM/) have goodprediction effect compared with RBF. 2. Aiming at aeration system transfer coefficient of dissolved oxygen in the gain controlparameter uncertainty, presents an adaptive neural control /(ANC/) strategy that guaranteesglobally asymptotic tracking for a class of uncertain nonlinear systems with function-typecontrol gains. A proportion differentiation /(PD/) control term with variable gain is developedto globally stabilize the plant so that neural network approximation is applicable. A statetransformation is applied to solve the control singularity problem resulting from the unknowncontrol gain function. A robust control term is developed to achieve asymptotic tracking of theclosed-loop system. Compared with previous global asymptotic tracking ANC approaches, theproposed approach not only simplifies the selection of PD gain, but also relaxes chattering atthe control input. Simulation results have demonstrated the effectiveness of the proposedapproach. The simulation experimental results show that the dynamic stability in this paper,the proposed method cans effectively the concentration of dissolved oxygen compared withPI. 3. Aiming at sludge concentration control recyle sludge discharge systems the direction ofthe uncertainty problem, presents a new robust adaptive fuzzy control algorithm directly。Theoverall control input contains a basic direct AFC term and an additional robust control term. ALyapunov-based ideal control law is proposed to solve the control singularity problem and theNussbaum gain technique is applied to solve the control direction problem. Using ane2-modification in adaptive laws, it not only obtains bounded adaptive parameters, but alsoachieves asymptotic convergence of tracking errors. Moreover, the proposed controller hasmore compact structure compared with the previous indirect approaches. Simulated studieshave demonstrated the effectiveness of the proposed approach.the simulation results show thatthis method is to be discharged through the recyle sludge, the sludge concentration of sludgeeffectively stabilize biochemical pool, verify the validity of the proposed method in this paper. 4. Aiming at the optimal control problem of wastewater treatment process operation cost,the fuzzy discrete particle swarm optimization control is proposed to calculate the optimalvalue of operation cost, which takes the two most important control parameters, sludgewastage and dissolved oxygen as control variables, regards total substrate discharge andeffluent water quality as restriction factors and operation cost of residual sludge treatment,sludge recyle and aeration as performance index. In the same water quality under the premise,based on the fuzzy particle swarm of inertia weight strategy to find the lowest cost of theoptimal solution. The test results show that the method can effectively improve the diversityof particle swarm, a reliable global convergence and fast convergence speed, can in thewastewater treatment process optimization effectively in dynamic environment. Simulation ofoptimal algorithm is of high search efficiency and low mean and variance of wastewatertreatment operation cost.

关 键 词： 污水处理系统 支持相量机 自适应神经网络控制 自适应模糊控制 模糊离散粒子群

分 类 号： [X703 TP273]

领　　域： [环境科学与工程] [自动化与计算机技术] [自动化与计算机技术]