【作者】 李胜海；

【导师】 郭劲松；

【作者基本信息】 重庆大学 ， 市政工程， 2002， 硕士

【摘要】 人工神经网络(ANN)和遗传算法(Genetic Algorithms简写为GA)是复杂非线性科学和人工智能科学的前沿，其在水污染控制领域的应用研究在国内外尚处于初创阶段。本文在较全面分析评述了水污染控制方法研究现状，及简要介绍ANN和GA的基本原理、优化算法的基础上，在国内首次将ANN方法中的离散Hopfield网络(Discrete Hopfield Neural Networks简写为DHNN)和GA引入水污染控制规划领域，主要在水污染控制方案优化方面进行了一些探索性的研究工作，为提高水污染控制方案优化水平做出努力。 本文根据水污染控制的特点及Hopfield网络能量函数在一定的条件下收敛于一定区域最低点的特点，通过对DHNN能量函数的合理设计，把水污染控制方案的费用转化为DHNN的能量函数，即使水污染控制方案费用最低的方案与DHNN能量函数的最低点相对应，提出了水污染控制方案优化DHNN模型，并从数学上严格推证了DHNN能量函数的收敛性。实例研究表明：DHNN得到的最优方案充分利用了自然水体的环境容量，而且，在选择优化方案的过程中，当某一控制断面的某种污染物超过规定的水质标准时，DHNN首先采用能减少该控制断面某污染物贡献单位浓度费用最低的方案组合，因而可以很快得到最优方案组合。 GA是一种具有普适性的数值求解方法，且对目标函数的性质几乎没有要求，甚至不要求显式地写出目标函数。GA用于优化计算时，不能直接处理问题空间的参数，而只能处理以基因链码形式表示的个体。本文巧妙地把遗传算法的基因链码同水污染控制方案相结合，即采用实数编码的形式，用实数1，2，3来分别表示各个污染源不处理、一级处理、二级处理的情况，同时，把个体的适应度函数同方案组合的总费用相对应，使得方案组合总费用最小的个体的适应度函数值最大，这样，由GA得到的优秀个体就对应于最优或逼近最优方案。实例研究表明：由于GA是随机选取初始值，且在算法上尚存在不足，从而导致GA得到的结果在最优解附近振荡。但是，由于GA在运行中始终保留一定数目的优良个体，因此，这些结果应当是最优或逼近最优结果的集合，即GA得到的结果均收敛于最优解的邻域内。决策者可以通过可行性、经济性、和科学性的比较从中找到最优方案。 本研究表明：用DHNN和GA来进行水污染控制方案优化在理论上可行，在实践上有继续深入研究开发的价值，具有良好的应用前景。本研究为水污染控制方案优化提供了一种新方法，开辟了一条较好的新途径，也为ANN和GA的应用增添了新领域。更多还原

【Abstract】 Innovative work is done on the applying of discrete hopfield neural networks (DHNN) and genetic algorithms (GA) to the optimization of the planning scheme of water pollution control in this dissertation.Artificial neural network and genetic algorithms play the leading role in the sciences for complex non-linear phenomena and artificial intelligence. Researches on its application in the water pollution control are still in the preliminary stage in the world. On the basis of a comprehensive evaluation and analysis of the present situation of the researches in the way of water pollution control, and on the basis of a careful exposition of the basic principles and the optimal algorithm of DHNN and GA, this dissertation gives an application of DHNN and GA approach in the scheme of water pollution control, which, as the first attempt of its kind, can help to achieve a higher level in the application of artificial intelligence in this field.Based on the features of integrated water pollution control and the astringency of the energy function of DHNN at certain condition, by the rational design of the energy function, the total cost of the scheme is transformed into the energy function of the DHNN, that is, the optimal scheme is corresponding to the point whose energy function is the lowest, the optimal model of DHNN of the water pollution control is put forwarded, and, at the same time, a strict mathematical deduction of the astringency of the energy function of DHNN is given. Case study reveals that: the optimal scheme of DHNN can make the use of environmental capacity sufficiently, and, when the concentration of a polluter in certain section gains over the aim in the optimizing process, the scheme that the cost of reducing per contribution concentration at this section is chosen, so it can get the optimal scheme quickly.Genetic algorithms is a widely used amount method, which almost has no requirement for the object function, even almost has no requirement for obvious object function. When it is used to optimizing calculation, it can’t deal with the matter’s parameter directly, but can only deal with the individual population in gene chain code. The gene chain code of the GA is combined to the scheme of water pollution control trickly, that is, using one, two, three to mean for the nontreatment, primary treatment and second treatment.At the same time, the total cost of the optimal scheme is corresponding to the adaptation function of GA,which corresponds the higheradaptation function to the optimal scheme, thus, the excellent individual is the optimal scheme. Case study reveals that: the result of the GA is unsteadiness, because the original is chosen freely, and the GA is finished by certain iterative time. Althogh the results are unsteadiness, they both are the neighborhood of the optimal. Decision-maker can get the best one through the comparison of the feasibility > the economical and the scientific.This research demonstrates that with its theoretical feasibility and great practical utility, the applying of DHNN and GA to the optimization of the scheme of water pollution control has good prospects for further development and application. This research proposes a new way of thinking for studies of he optimization of the scheme of water pollution control and adds to the fields of application of ANN.更多还原