遗传算法改进神经网络模型在太子河流域径流预报中的应用

本文对用遗传算法(GA)改进的神经网络的算法进行了分析,建立了用GA寻求权重的前馈网络模型(简称模型I)和用GA优化初始权重的前馈网络模型(简称模型Ⅱ)。对太子河流域的5个计算单元前35年年径流量进行了模拟和用后10年资料对模型进行了检验,并与传统的BP神经网络模型进行了比较。结果表明:模型I对历史样本拟合较差,但其预报效果优于模型II和传统的网络模型。模型II和未改进的网络模型效果相似,精度有所提高。     

基于BP神经网络和遗传算法的丰满水库洪水预报模型研究

根据丰满水库的流域特性,尝试使用BP神经网络模型进行洪水预报。针对随机生成网络权重的盲目性,采用遗传算法进行初始权重优化。通过历史洪水检验,证实此模型在丰满水库的洪水预报中有很高的预报精度和应用价值。     

人工神经网络与遗传算法在河道洪水预报中的应用

在总结大量洪水智能预报实践经验的基础上，主要作了两方面的工作：（1）提出了一种能够进行峰值识别的改进BP算法（Error Back Propagation with Peak Recognizer，简称BPPR），该算法在个性网络权重时偏重大值，即大值误差对权重的个性起主要作用。从而提高模型对洪峰峰值的预报精度。（2）采用遗传算法优化网络的初始权重，以解决BP网络训练容易落入局部极小点的问题。     

基于神经网络的小型水库洪水预报研究

针对经验预测方法精度不高,传统水文模型应用至小型水库进行洪水预报工作量大、推广较难的问题,引入具有强大特征学习能力的人工神经网络(ANN)方法,结合遗传算法(GA)寻参,对小型水库进行洪水预报。利用GA实现ANN中时间步长和隐含层神经元节点参数自动寻优,可避免寻参盲目性,针对性地为各小型水库构建个性化洪水预报模型。通过构建反向传播(BP)、长短期记忆(LSTM)、门控循环单元(GRU)神经网络洪水预报模型,对实测洪水过程进行模拟对比试验。结果表明：LSTM模型预报精度高、稳定性良好,能学习并模拟实际洪水过程水位变化规律,预报性能优于BP和GRU模型。     

基于贝叶斯平均法的洪水分类组合预报研究

以实测样本隶属于某一模型的后验概率为权重,对各模型预报变量的后验分布进行加权平均,获得综合预报变量的概率密度函数,进而推导出均值、方差公式和置信区间。将“基于BP神经网络和遗传算法的分类洪水预报”与“基于贝叶斯模型平均法的多模型组合预报”相结合,实现“洪水分类多模型加权组合预报”。该方法综合考虑了各模型对不同类型洪水的适应条件及多个模型的优势,可同步降低模型参数与模型结构不确定性对预报结果的影响。以大伙房水库为研究区域,采用大伙房水库1960年~2016年的25场历史典型洪水作为本研究的基础资料,分别完成实例验证与分析,效果整体上优于单个模型。     

基于神经网络与遗传算法的洪水分类预报研究

针对水文模型参数的不确定性,对洪水进行分类预报,不同类型洪水采用不同预报参数,旨在提高洪水预报精度。基于BP神经网络模型,依据分类因子选取原则,选取6项具有代表性的影响因子作为模型输入,可将洪水划分成高、中、低3类。基于遗传算法,对3类洪水进行参数率定,获得3组不同的参数组,最终利用训练好的分类预报模型实现不同类型洪水的变参数预报。以大伙房水库25场历史典型洪水进行实例验证与分析,结果表明:分类预报结果的洪峰误差、峰现误差、确定性系数及典型洪水过程的拟合效果明显优于分类前。经训练后的基于BP神经网络与遗传算法的洪水分类预报模型可较好适用于大伙房水库,结果更贴合实测值,效果整体上优于分类前,方法可行、有效。     

小浪底-花园口区间洪水智能预报方法研究

结合实际资料,以改进的神经网络模型检验了小浪底一花园口区间洪水智能预报方法的可行性和精度的可靠性,并对出现个别较大误差的原因进行了分析。研究结果表明,采用遗传算法优化网络初始权再和引人峰值修正系数的改进BP算法进行洪水预报,其模型合理、算法可靠、精度较高。同时,为充分发挥神经网络对复朵系统映射的优势,建议：①加强水文观测力度,获得较为准确的观测资料,包括连续的水位、流量资料,洪峰传播时间以硬区间人流等;②在实际应用中直接采用当前洪水数据作为训练样本,以使网络更能反映当前洪水的演进规律,并提高预报精度;③开展误差实时修正的研究,将误差修正方法引入到模型中;④扩大建模范围,建立降雨-径流预报模犁,并考虑区间入流情况,以进一步提高网络的预报精度,延长预见期。     

改进Elman神经网络在径流预测中的应用

针对传统静态前馈神经网络动态性能较差的缺点,提出一种基于遗传算法（GA）优化Elman神经网络连接权值的GA-Elman多元变量年径流预测模型.以新疆伊犁河雅马渡站径流预测为例进行实例分析,并构建传统Elman,传统BP和GA-BP多元变量年径流预测模型作为对比模型,预测结果与文献IEA-BP网络模型预测结果进行对比.结果表明:①GA-Elman模型的拟合及预测效果略优于文献IEA-BP模型,该模型用于多元变量年径流预测是合理可行的,具有较好的预测精度和泛化能力.②在相同网络结构及传递函数等条件下,GA-Elman模型的预测精度和泛化能力优于GA-BP模型,传统Elman模型优于传统BP模型,表明具有适应时变特性的Elman反馈动态递归网络预测性能优于BP网络;GA能有效优化Elman神经网络连接权值,使网络的预测精度和泛化能力有了较大提高.     

人工神经网络与遗传算法在多泥沙洪水预报中的应用

由于水沙作用机制和演进规律的复杂性，以及河道形态变化等因素，多泥沙洪水预报一直是洪水预报的难点，对高含沙洪水快速、准确的预报是多年来国内外专家十分关注的课题。作者采用具有高度非线性识别能力的人工神经网络与遗传算法相结合的方法，探讨了建立智能预报模型的基本方法，进一步对如何提高预报精度的问题进行了研究，并结合黄河洪水预报实例检验了神经网络模型的可行性。检验结果表明，该方法能够较好地识别多泥沙洪水的演进规律，对水位、流量和含沙量都能进行合理预报。     

基于遗传神经网络的河流冰凌预报

在分析影响凌汛成因因素的基础上,选取合适的预报因子,建立了基于遗传算法(GA)和Levenberg-Marquardt BP网络(LMBP)相结合的冰凌预报神经网络模型.GA-LMBP算法先通过遗传学习算法进行全局训练,再用LMBP算法进行精确训练,克服了传统BP算法收敛速度慢、易陷入局部极小等缺陷.把该模型应用到松花江依兰、佳木斯段开河预报中,取得了良好的效果.     

地下厂房施工过程的三维非线性数值模拟分析

根据冗各电站地下厂房开挖支护的施工顺序,以隐式杆单元模拟锚杆,以壳单元模拟混凝土喷层,采用三维非线性有限元技术,对洞室群的开挖支护动态过程进行了模拟计算。通过三维非线性有限元计算分析得到:洞室群动态开挖过程中围岩应力变形的分布规律,围岩塑性区的发展规律,锚杆受力变化情况。计算结果表明:整个开挖过程中,洞周围岩位移变化规律正常,量值较小,塑性区主要分布在洞室周边,围岩稳定性较好;但在洞室交叉口处围岩的变形较大,出现较大范围的塑性区,在此基础上提出了支护措施优化方案,并对优化方案的效果进行了评价。     

Evaluation of freshwater mussel relocation as a conservation and management strategy

The relocation of unionacean mussels is commonly used as a conservation and management tool in large rivers and streams. Relocation has been used to recolonize areas where mussel populations have been eliminated by prior pollution events, to remove mussels from construction zones and to re-establish populations of endangered species. More recently, relocation has been used to protect native freshwater mussels from colonization by the exotic zebra mussel Dreissena polymorpha. We conducted a literature review of mussel relocations and evaluated their relative success as a conservation and management strategy. We found that 43% of all relocations were conducted because of construction projects that were forced to comply with the Endangered Species Act 1973 and that only 16% were monitored for five or more consecutive years. Most (43%) relocation projects were conducted from July to September, presumably a period when reproductive stress is relatively low for most species and the metabolic rate is sufficient for reburrowing in the substrate. The mortality of relocated mussels was unreported in 27% of projects; reported mortality varied widely among projects and species and was difficult to assess. The mean mortality of relocated mussels was 49% based on an average recovery rate of 43%. There is little guidance on the methods for relocation or for monitoring the subsequent long-term status of relocated mussels. Based on this evaluation, research is needed to develop criteria for selecting a suitable relocation site and to establish appropriate methods and guidelines for conducting relocation projects.     

基于模型试验的河床砂砾石层基本特性研究

为了经济可靠地确定深厚覆盖层的基本特性参数,以大渡河双江口水电站坝基河床砂砾石层基本特性研究为工程背景,进行大尺寸模型试验,研究河床砂砾石层的旁压模量和动探击数随密度、级配以及上覆压力的变化规律。根据室内模型试验结果,结合现场旁压试验及钻孔试验成果,推断出坝址区河床砂砾石相关层位的密度,并依此进行砂砾石室内力学试验,获得设计所需力学参数。研究成果可为深厚覆盖层基本特性测试提供新的方法。     

基于BP神经网络-马尔科夫链模型的隧道围岩位移预测

在隧道工程施工中,围岩位移预测起着很重要的作用。将BP神经网络-马尔科夫链模型引入到隧道围岩位移预测中来,通过对训练样本的学习,利用BP神经网络实现了对位移时间序列的滚动预测,同时得到了实测值与预测值的相对误差;在此基础上利用马尔科夫链对相对误差进行修正,有效地提高了预测结果的精度。并将该模型应用于某公路隧道拱顶下沉位移时序预测中,结果表明该模型具有精度高、科学可靠的特点,为隧道围岩变形的预测提供了新的途径。     

Short communication. Pipelines as a refuge for freshwater snails

In the lower River Murray, Australia, populations of gastropods have declined markedly over the past 40 years. Of the native fauna of about 18 species, only the ancylid Ferrissia petterdi remains common (but in patches), and another five species feature in sporadic local resurgences. Until recently the viviparid Notopala was thought to be extinct. The introduced physid Physa acuta is widespread, but in low numbers. The decline probably is associated with flow regulation, high turbidity and the introduction of common carp. Snail populations have persisted, however, in irrigation pipelines fed from the river, where they periodically cause blockages. This paper reports the discovery of the only known populaion of Notopala hanleyi in a pipeline where the inner wall accumulates microbial and other organic matter, providing high quality food for snails. In the Murray before regulation most of the periphytic biomass was probably microbial, as fluctuating water levels would have maintained littoral communities in a highly productive, early state of succession his postulated that, by stabilizing seasonal water levels, regulation has promoted the production of filamentous green algae, an inferior food source, and so displaced snail populations from the open aquatic environment.     

浙江省水域调查录入系统的设计和应用

浙江省水域调查录入系统将调查数据与电子地图相整合,对分类水域数据进行统计计算,并对输入数据进行合理性和完整性检查,得到调查范围内现有水域的详细数据,实现了水域调查数据输入、计算、统计和查询的数字化管理,提高了工作效率和工作质量,为各地下阶段水域保护规划编制和水域管理工作提供了数据支持。对该系统的设计原则、功能和技术设计等作了介绍。     

基于CT-连续加荷水泥乳化沥青混凝土细观损伤规律研究

利用CT技术对水泥乳化沥青混凝土的细观损伤过程进行实时扫描观测,获得了混凝土试件在连续加荷下的CT图像。通过采用直观分析法、CT数分析法、CT数与损伤变量关系3种方法分析表明,水泥乳化沥青混凝土的破坏过程可分为压密、扩容、裂纹扩展、破坏4个阶段,同时混凝土在受压过程中具有韧性,在极限荷载之前各断面处于压密和微扩容的强化阶段,极限荷载之后混凝土从损伤到破坏发展比较平稳。     

Automated system for removal and pneumatic transport of fly ash from electric precipitator hoppers

A system for removal and pneumatic transport of fly ash is examined, in which air pulses act on batches (pistons) of ash formed in a duct. Studies are made of the effect of several physical parameters on the force required to displace a piston of ash and these serve as a basis for choosing a system for removal and pneumatic transport of ash simultaneously from several hoppers of an electric precipitator. This makes it possible to separate the ash particles according to size without introducing additional components. Formulas are given for calculating the structural and dynamic parameters of this system and measurements of indirect dynamic parameters are used to calculate the input-output characteristics of the system. In order to optimize the system, configurations for summing several ducts into a single transport duct for pneumatic ash transport are proposed. Some variants of dry ash utilization and the advantages of producing of size-separated particles are considered. __________ Translated from élektricheskie Stantsii, No. 2, pp. 26–30 (2008).     

The flood pulse advantage and the restoration of river-floodplain systems

The ‘flood pulse advantage’ is the amount by which fish yield per unit mean water area is increased by a natural, predictable flood pulse. Evidence for this increase is presented from tropical and temperate fisheries. It is argued that increasing multispecies fish yield by restoring the natural hydrological regime is consistent with increasing production of other trophic levels and with restoration from ecological and aesthetic viewpoints. When applied to a river-floodplain system, this restoration would provide a large, self-sustaining potential for recreation, commercial exploitation, and flood control. An interim ‘natural flood pulse’ restoration approach is proposed for systems modified for navigation. This approach approximates the natural hydrological regime in a river reach and is intended as a first step in the long process of restoring the watershed.     

浅谈天湖水电站水轮机运行故障及处理措施

简述天湖水电站投运以来各种故障的表现形式，分析其原因，提出了相应的处理方法及措施，并对合理利用国内外设备提出建议。