学位论文-—环境建模与软件外文翻译

1、外文翻译Environmental Modelling&Software1.1. BackgroundEnvironmental change, economic and social pressures, and limited resources motivate systems analysis techniques that can help planners determine new management strategies, develop better designs and operational regimes, improve and calibrate simulat

2、ion models, and resolve conflicts between divergent stakeholders. Metaheuristics are emerging as popular tools to facilitate these tasks, and in the field of water resources, they have been used extensively for a variety of purposes (e.g. model calibration, the planning, design and operation of wate

3、r resources systems etc.) in many different application areas over the last few decades (Nicklow et al., 2010). Since metaheuristics were first applied in the water resourcesfield (Dougherty and Marryott,1991; McKinney and Lin,1994; Ritzel et al.,1994; Gupta et al.,1998), their popularity has increa

4、sed dramatically, probably facilitated by the simultaneous increase of available computational power (Washington et al., 2009), to the point where they are widely used (Nicklow et al., 2010), even by actual water planning utilities (Basdekas, 2014).Zufferey (2012) defines a metaheuristic“as an itera

5、tive generation process whichguides a subordinate heuristic by combining intelligently different concepts for exploring and exploiting the search space”, as part of which“ learning strategies aused to structure information in order to find efficiently near-optimal solutions. ” Unlike more “ traditio

6、nal app”roaches, which use mathematical programming to specify the optimal value of one or more objective functions, metaheuristics incorporate elements of structured randomness for search and follow empirical guidelines, often motivated by observations of natural phenomena (Collette and Siarry, 200

7、3).Metaheuristics can be divided into two groups, including population-based algorithms (e.g. genetic algorithms, evolutionary strategies, particle swarm optimization, ant colony optimization, etc.) and single point-based methods (e.g. simulated annealing, tabu search, simple (1+1) evolutionary stra

8、tegies, trajectoryor local search methods, etc.). Evolutionary algorithms (EAs) are the most well-established class of metaheuristics for solving water resources problems and are inspired by various mechanisms of biological evolution (e.g. reproduction, mutation, crossover, selection, etc.) (Nicklow

9、 et al., 2010). Consequently, the focus of the remainder of this paper is on EAs, although many of the concepts discussed also broadly apply to other metaheuristics. The paper also provides general guidelines and future researchdirections for the broader class of systems analysis approachesthat take

10、 any sort of optimisation into account.When using EAs, the steps in the optimisation process generally include:1. Problem formulation (i.e. selection and definition of decision variables, objectives, and constraints).2.Selection of decision variable values.3. Evaluation of objectives and constraints

11、 for the selected decision variable values, which is generally done using one or more simulation models.4.Selection of an updated set of decision variable values based on feedback received from the evaluation process using some search methodology.5.Repetition of points 3 and 4 until the selected sto

12、pping criterion has been satisfied.6.Passing the optimal solutions into an appropriate decision-making process.As outlined below, compared with more “ traditional opti”misation methods, EAs have a number of advantages, which are most likely responsible for their widespread adoption for water resourc

13、es problems.1. The basic analogies that inform their optimisation strategies are conceptually easy to understand.2. As simulation models are generally used to calculate objective function values and check constraints, it is easy to add optimisation to existing simulation approaches. This gives rise

14、to the potential for greater confidence in the results by end users, as the outcomes of the optimisation process are based on the results of simulation tools that are already used for the purposes of decision-making.3. EAs are capable of solving problems with difficult mathematical properties (Reed

15、et al., 2013). This is because the ability to link with simulation models reduces the need for problem simplification, which is required for many traditional optimisation algorithms that are unable to deal with nonlinearities (e.g.exact finitely terminating algorithms, like linear and nonlinear prog

16、ramming) or discontinuities (e.g.iterative/convergent algorithms, such as first or second order gradient methods). For example, in linear programming applications, there is no ability to account for nonlinearities, such as -then ”“ isf tyle rules.Consequently, the philosophy underpinning EAs is that

17、 it is generally better to find near globally optimal solutions to the actual problem, rather than globally optimal solutions to a simplified problem, especially when the simplified problem misses key socially relevant properties(RittelandWebber,1973).4. The linking with simulation models facilitate

18、s the straight forward treatment of parallel computing.5. EAs have the ability to perform both exploration (i.e. global search) and exploitation (i.e. local search) of the fitness function, increasing the chances of finding near-optimal solutions to complex problems (Nanda and Panda, 2014).6. The al

19、gorithms themselves are readily adaptable to a wide variety of application contexts (Back et al., 1997; Goldberg, 1989;Nicklow et al., 2010).1.2. Purpose and organisation of this position paperThis position paper aims to contribute to the literature that reviews EA algorithms (e.g. Coello et al., 20

20、07; Deb, 2001) and EA use in water resources (e.g.Nicklow et al., 2010), and performs diagnostic assessments on water-related problems (e.g. Reed et al., 2013). However, the primary purpose of this paper is to map out the most important research challenges and future directions in applying EAs to th

21、e complex, real-world water resource applications that are most in need of these methods.While a brief review of current progress in relevant areas is provided, it should be noted that this is not meant to be a comprehensive review paper.The research challenges identified in this paper are motivated

22、 by the fact that over the last 20e25 years, much of the research in the field of EAs in water resources has focused on the application of different types of algorithms to different problem types. In the majority of these studies, the aim was either to :1. Develop and test the performance of differe

23、nt types of algorithms (e.g. an algorithm inspired by a different natural phenomenon (e.g. genetics, the foraging behaviour of ants in search for food, the behaviour of bees and birds etc.), variants of existing algorithms, or hybrid algorithms; or 2. Test if a particular algorithm or variant can be

24、 used successfully to solve different types of water resources optimisation problems (e.g. model calibration,water distribution system design.groundwater remediation, environmental flow allocation, reservoir operation etc.) and/or different instances of these problem types (e.g. using different mode

25、ls, in different geographical locations etc.).译文环境建模与软件1.1 背景环境变化， 经济和社会压力， 以及有限的资源激励系统的分析技术， 可以帮 助规划者确定新的管理策略， 开发出更好的设计和运作机制， 完善和校准仿真模 型，并解决分歧之间的冲突利益相关者。 共通启发式演算法正在成为流行的工具， 以促进这些任务的完成，并在水资源的领域中，它们已被广泛地用于各种目的 （例 如模型校准， 水资源系统的规划， 设计和操作等） 在过去的几十年中的许多不同 应用领域（ Nicklow 等人， 2010）。因为共通启发式演算法是首先被应用在在水 资源领域（

26、 Dougherty 和 Marryott ， 1991; McKinney 和 Lin ，1994; Ritzel 等 人， 1994; Gupta 等人， 1998），它们的知名度显著增加，可能是可用的计算能 力的同时增加便利（ Washington 等， 2009），指向在广泛使用的地方（ Nicklow 等， 2010），即使按实际用水规划公用事业（ Basdekas 2014）。Zufferey （2012）定义了一个共通启发式演算法“作为迭代产生的过程，通 过组合对勘探和开发的搜索空间智能不同的概念引导下属启发式” ，其中“学习 策略用于结构信息，以便有效地找到接近最优的解决方案。

27、不同于更“传统的” 方法，它使用数学规划指定一个或更多个目标函数的最佳值， 共通启发式演算法 结合结构随机性的搜索元素， 并按照经验准则， 通常通过自然现象的观测动机 （科 莱特和 Siarry ， 2003）。共通启发式演算法可分为两类， 包括人口为基础的算法 （如遗传算法， 进化 策略，粒子群算法， 蚁群算法等）和单点为基础的方法 （如模拟退火， 禁忌搜索， 简单（1+1）进化策略，轨迹或本地搜索的方法等）。进化算法（EAS是在解决 水资源问题上最行之有效的共通启发式演算法， 并通过生物进化的各种机制 （例 如复制，变异，交叉，选择等）的启发（ Nicklow 等， 2010）。因此，本文

28、的其 余部分的重点是进化算法， 虽然很多被讨论的概念还广泛适用于其他共通启发式 演算法。本文还提供了一般准则和未来的研究方向， 为更广泛的种类是采取任何 形式的优化考虑系统的分析方法。当使用进化算法，在优化过程中的步骤一般包括：1. 问题描述（即选择和决策变量、目标和约束的定义）2. 选择决策变量的值。3. 评价的目标和约束的决定所选变量的值， 这一般是使用一个或多个仿真模 型。4. 选择更新组的决策变量值的基础使用一些搜索方法的评价过程从收到的 反馈。5. 重复点 3 和 4 ，直到满足所选的停止标准满足。6. 进入适当的决策过程中的最优解。 如下面所述，与更“传统”的优化方法相比，进化算法具有许多优点，其中 最有可能造成他们对水资源问题的广泛采用。1. 通知其优化策略的基本类比是在概念上容易理解。2. 作为仿真模型通常用于计算目标函数值， 并检查约束，很容易添加优化现 有模拟方法。 这产生了潜在的来自终端用户的更大的信心， 作为优化过程的结果 是基于对模拟工具已经用于决策的目的的结果。3. 进化算法能够解决困难的数学特性的问题