基于深度强化学习算法的高能效数据负载均衡方法

网络用户大幅度增加为网络发展带来机遇的同时也带来了挑战,当前使用的数据负载均衡方法节点数据分发能力较差导致网络节点吞吐量较低。因此,设计了基于深度强化学习算法的高能效数据负载均衡方法。选择隐式并行程序设计方法,设计网络数据并行程序。根据存储节点个数,设计数据分配与迁移方法。根据存储节点数据量与特征,选择深度强化学习算法,实现高能效数据负载均衡。构建仿真分析环节,经多次测试证实,深度强化学习负载均衡方法具有调节网络节点平均吞吐量的能力,且此方法的使用效果优于当前所使用的其他方法。     

Hadoop负载树任务调度算法

针对现有异构任务调度算法存在负载不均衡、数据本地性问题,提出基于树结构的负载树任务调度算法。该算法通过量化节点计算能力构造节点集最小堆,利用堆排序生成计算能力逆序树,并依据节点负载率将逆序树调整为左节点优先的负载树,为任务计算包含完成时间、负载率、延迟因子的决策值,最终完成任务与树节点的匹配。实验结果表明,取不同负载率与延迟权值比时,该算法的任务执行效率均能获得一定程度提高。该算法可利用树结构的调度优势,在获得更高集群负载均衡度时,有效缩短作业集执行时间。     

覆盖多播节点失效检测的分布式算法的改进

覆盖多播在端系统间构建数据分发树．由于端系统随意地加入或退出多播组,其可靠性不如路由器．当分发树中的非叶节点退出或者节点失效时,其下游节点将受到影响．低代价而快速的节点失效检测能降低由于服务间断所带来的影响,同时也为后续的恢复工作提供一个良好的基础．在分析相关工作的基础上,提出了一种改进的节点失效检测机制,它能够兼顾节点失效检测时间和额外消息负载两个性能指标．     

一种基于共享机制的最小通讯负载平衡算法

为有效减少在分布式并行交通仿真中仿真车辆在主机间跃迁、交换信息等引起的超高的主机开销,提出了一个基于共享机制下的任务划分方法,采用了具有自适应能力的负载均衡预测模型进行模拟实验.实验表明该方法具有占用资源少、任务划分相对均衡、主机间通信相对较少等特点,基本实现了节点间最小通讯的负载均衡的目的.     

基于MPI的动态负载均衡任务分配方法设计实现

在拥有不同计算能力的异构机群中,并行计算任务调度问题是一个NP-完全问题,现有的任务分配处理方法中．对机群任务分配与监控处理以及对机群节点计算能力和负载方面考虑较少,在实际计算中可能严重影响机群性能,降低计算效率,而在MPT中引入动态负载均衡的办法能够提高任务分配质量和计算节点工作效率,实验结果表明动态负载均衡模型能够很好地解决任务调度的问题。     

云环境下聚类分解的高维数据混合索引方法

针对云计算环境下分布式存储系统的数据索引不支持复杂查询的问题,笔者提出了云环境下聚类分解的高维数据混合索引方法.首先,采用聚类分解方法对分割数据建立树状索引;然后,以叶节点为单位,通过扫描线算法来获取节点内部所有对象的局部最近邻结果;最后,依据计算的结果得出启发式的裁剪距离.在单节点最近邻计算中,第二个阶段获取外部的最近邻对象采用范围查询算法.实验分析表明,在查询效率上该索引方法高于单纯的聚类方法.与M-tree、顺序查找、iDisance相比,基于聚类分解的混合索引方法在高维查询模式下具有良好的查询效率和负载均衡.     

基三分层互连网络中基于受限多播树的动态负载平衡

为了解决基三分层互连网络(THIN)系统中的负载平衡问题,提出一种采用多播树技术提高节点间交换负载信息效率的动态负载平衡(DLB)算法--THINDLBA.设计了一套完整的DLB消息和各节点处的信息维护机制以辅助算法实现.重载节点的负载迁移请求消息沿着一棵以该节点为根的多播树传播,被该树覆盖的轻载节点均成为负载迁移的候选目标节点,可以沿着该树和重载节点交互负载信息,从而使重载节点能够在算法的一次执行中外迁最多的过载进程,尽快改善自身负载状态.算法设计中约束了多播树的构造过程,以避免因树间覆盖造成的消息误传或冗余.通过实验对比了4种DLB算法的性能,结果证明THINDLBA能更有效地缩减THIN系统处理计算密集型任务的时间.     

基于生成树的WSN溯源数据压缩方法

在无线传感器网络(WSN)中溯源数据(Provenance)记录了一个数据从产生至被传输到基站(BS)途经的所有节点以及在这些节点上对数据的操作。提出一种基于生成树的溯源数据压缩方法,其基本思想是在字典中存放WSN拓扑图的生成树并对其建立索引,在数据包传输过程中传输的是生成树的索引而不是完整的生成树。仿真实验结果表明,在大规模稀疏WSN中采用该方法,溯源数据在文件大小和传输能耗等方面都优于已知的其它溯源数据编码技术,而且该方法对线性溯源数据和聚合溯源数据采用完全相同的处理方式,算法实现简单、一致性好。     

具有前向安全性的组密钥分发方案设计与实验分析

针对已有组密钥分发方案中的安全隐患,基于对称二元多项式提出一种无需可信第三方的具有前向安全性的组密钥分发方案。方案中,组内每对用户利用对称二元多项式的特征协商出一个对称密钥,实现组密钥的安全分发。每个用户维护一张通信记录表用于更新用户间的对称密钥,实现组密钥分发的前向安全性。针对方案的正确性、安全性和性能优势,给出了详细的证明和仿真实验分析。     

e-Learning Grid节点切割负载均衡机制研究

大规模并发用户环境下的e-Learning教育应用,必然要受硬件、网络和其他资源的约束和限制,严重时可导致应用系统的整体瘫痪.以节点切割为代表的负载均衡机制,可在一定程度上缓和无限增长的并发用户给系统有限资源(硬件、软件、网络等)带来的压力,提高应用系统的可靠性、稳定性和可扩展性.本文以e-Learning Grid数据访问中间件为基础,详细讨论了节点切割的原理、步骤、实现和示范应用,给出了以负载均衡为目标的负载转移服务的完整实现策略,并通过测试实验分析了其对系统性能的提升.     

机械零件随机疲劳载荷的统计分析方法

研究了机械零件载荷的时间历程特性，分析了随机载荷的循环计数法、峰值计数法和雨流计数法等统计分析方法，提出了随机疲劳载荷的概率分布函数及疲劳载荷谱的编制方法。     

Contact analysis and load distribution of double-envelop hourglass worm gearing affected by errors and load

1. Introduction The double-envelop hourglass worm gearing is characterized by multi-tooth line contact and good lubrication condition. It has high load-carrying capacity and is often built in heavy-duty machines. Like other line-contact gear drive, it is sensitive to manufacturing errors, misalignment errors and load-dependent deformations, which cause the tooth end contact and large transmission errors [1]. The classic theory for meshing analysis of a gear drive overlooks diverse errors and l…     

一种新型的配电网线损管理及无功优化系统

介绍一套集配电网络的图形编辑、线损计算分析、无功优化以及报表管理一体化的系统,可基于图形和SCA-DA系统实时数据共享,接入配变监测、负荷控制系统,整合成完整的线路或台区数据,包括设备属性以及各种实时运行数据等.采用改进强引导粒子群与混沌优化相结合的算法进行无功优化.该算法解决传统粒子群寻优后期粒子可能陷入早熟收敛的问题.通过对福建某高压配电网的具体计算,最优降损率可以达到14.04%,节点最低电压从0.895 0（p.u.）提高到0.995 6（p.u.）,系统通过一种新型线损管理及无功优化模式,为工作人员提供一个决策支持的技术平台.     

Modeling of D-STATCOM in distribution systems load flow

This paper presents modeling of Distribution STATCOM (D-STATCOM) in load flow calculations for the steady- state voltage compensation. An accurate model for D-STATCOM is derived to use in load flow calculations. The rating of this device as well as the direction of required reactive power injection for voltage compensation in the desired value (1 p.u.) is de- rived and discussed analytically and mathematically by the phasor diagram method. Furthermore, an efficient method for node and line identification used in load flow calculations is presented. The validity of the proposed model is examined by using two standard distribution systems consisting of 33 and 69 nodes, respectively. The best location of D-STATCOM for under voltage problem mitigation approach in the distribution networks is determined. The results validate the proposed model for D- STATCOM in large distribution systems.     

Three-dimensional elasto-plastic finite element analysis of a soil-pile-bridge interaction

The soil-pile-bridge interaction of super-large pile groups is a very complex issue for the design of deep pile group foundations. In this paper, the load distribution on the pile top of a super large bridge foundation and its influential factors are analyzed comprehensively using a three-dimensional elasto-plastic finite element method. The adopted model and its input parameters are firstly verified by comparing the numerical results with the measured data of static loading tests of a single pile. Numerical analysis is then performed to investigate the load distribution and the load-settlement characteristics of super-large pile groups, and the models are verified using centrifuge laboratory model testing data. The mechanism of the interaction between pile groups and soil under different conditions is explored.     

Adaptive load forecasting of the Hellenic electric grid

Designers are required to plan for future expansion and also to estimate the grid＇s future utilization. This means that an effective modeling and forecasting technique, which will use efficiently the information contained in the available data, is required, so that important data properties can be extracted and projected into the future. This study proposes an adaptive method based on the multi-model partitioning algorithm （MMPA）, for short-term electricity load forecasting using real data. The grid＇s utilization is initially modeled using a multiplicative seasonal ARIMA （autoregressive integrated moving average） model. The proposed method uses past data to learn and model the normal periodic behavior of the electric grid. Either ARMA （autoregressive moving average） or state-space models can be used for the load pattern modeling. Load anomalies such as unexpected peaks that may appear during the summer or unexpected faults （blackouts） are also modeled. If the load pattern does not match the normal behavior of the load, an anomaly is detected and, furthermore, when the pattern matches a known case of anomaly, the type of anomaly is identified. Real data were used and real cases were tested based on the measurement loads of the Hellenic Public Power Cooperation S.A., Athens, Greece. The applied adaptive multi-model filtering algorithm identifies successfully both normal periodic behavior and any unusual activity of the electric grid. The performance of the proposed method is also compared to that produced by the ARIMA model.     

Analysis modeling for plate buckling load of vibration test

In view of the recent technological development, the pursuit of safe high-precision structural designs has been the goal of most structural designers. To bridge the gap between the construction theories and the actual construction techniques, safety factors are adopted for designing the strength loading of structural members. If safety factors are too conservative, the extra building materials necessary will result in high construction cost. Thus, there has been a tendency in the construction field to derive a precise buckling load analysis model of member in order to establish accurate safety factors. A numerical analysis model, using modal analysis to acquire the dynamic function calculated by dynamic parameter to get the buckling load of member, is proposed in this paper. The fixed and simple supports around the circular plate are analyzed by this proposed method. And then, the Monte Carlo method and the normal distribution method are used for random sampling and measuring errors of numerical simulati     

Active control experiments on a herringbone ribbed cable dome

Active control experiments on a newly proposed herringbone ribbed cable dome are described in this study. The cables of the dome are designed to have the ability to change length in order to adjust the geometrical configuration and the force distribution of the structure. Thereby, the dome is adaptable to different load cases. To begin with, for achieving the control amount for the active control test, an active control algorithm based on a nonlinear force method is presented. Then, an assembly and pre-stressing procedure is implemented. Active adjustment tests on three possible types of adjustable cables are performed to provide a practical method for the following active control test. The active control test demonstrates the applicability of the active control algorithm to achieve both force control and shape control. The method can be used to prevent failure of the cable domes due to slackening of the ridge cables and excessive displacements of the central section of the cable dome. The experiments verify the proposed control algorithm and the feasibility of the cable dome to adapt to excessive full span load and maintain the integrity of the structure.     

A novel control strategy for load converter of DC isolated distribution system under unbalanced loading conditions

A novel control strategy for the load converter supplying the unbalanced AC load in a DC isolated distribution system is presented. The control algorithm results in balanced and sinusoidal load voltages under unbalanced AC loading. The unbalanced load is characterized in the d-q-0 rotating coordinate based on symmetrical sequence components. Also, the mathematical model of the load converter in both a-b-c and d-q-0 coordinates is derived by using the average large signal model. Then, two control strategies for the load converter are presented. The first one uses the conventional d-q-0 controller to ensure the voltage and current regulation. The second one is a newly proposed control strategy based on the decomposition of the voltage and current into in- stantaneous positive, negative, and zero sequences. These three sequences are controlled independently in their own reference frames as DC signals. The performance of the load converter using these two control strategies is compared. Simulation results show the validity and capability of the newly proposed control strategy.     

基于MEEMD加窗改进方法与GRNN组合的电力负荷预测

针对电力负荷序列不稳定性、随机性等特点引起的电力负荷预测精度下降等问题,提出MEEMD加窗改进方法和GRNN组合的短期电力负荷预测模型。利用GRNN神经网络延拓方法对原始信号两端数据进行延拓,用余弦窗函数对延拓数据加窗处理后再进行MEEMD分解,用神经网络对各分量趋势进行预测,叠加各分量的预测结果得到负荷序列的最终预测结果。实验结果表明,MEEMD加窗改进分解预测的平均绝对误差、平均绝对值百分比误差和均方根误差分别为73.926 8、0.818 0%和82.930 1。基于MEEMD加窗改进方法和GRNN组合的电力负荷预测不仅能抑制端点效应,而且能解决模态混叠和伪分解问题,提高了短期电力负荷的预测精度。