黄河内蒙古段冰情的自适应神经网络预报

运用数理统计理论的互相关分析方法,对黄河内蒙古段冰情信息时间序列进行分析,筛选出了冰情信息的预报因子,建立了冰情预报的自适应BP神经网络模型。预报结果表明:流凌日期为乙等预报方案,封河日期为甲等预报方案,开河日期为丙等预报方案。     

年均径流预测的遗传神经网络模型研究

为了进一步提高BP网络模型对径流预测的精度,采用遗传算法优化了BP网络初始的权值和阈值。实例研究结果表明:该方法克服了传统BP网络极易陷入局部极小值点等缺点,提出的遗传神经网络预测模型能够提高预测精度。     

基于广义回归神经网络的无黏性土管涌判定研究

分析了现在广泛采用的判定管涌破坏手段的不足之处。在分析广义回归神经网络的基本原理和算法基础上,建立了无黏性土管涌判别的广义回归神经网络模型。以前人试验结果作为对比,采用特征粒径和孔隙率作为判别指标,对土样的渗透破坏形式进行判别。计算结果表明,该模型的管涌渗流破坏形式判定结果与前人试验结果完全一致,该方法为无黏性土管涌渗流破坏形式的判别提供了新的研究思路。     

基于神经元PID的风力发电机组独立变桨控制

依据风力机空气动力学原理,设计了神经元PID控制器,实现了风力发电机组3个桨叶的独立变桨控制。通过神经元PID控制,计算出桨距角,并与Park变换得到的附加桨距角相加,作为独立桨距角的设定值。基于Fast软件平台,以2 MW变速变桨风力发电机组为验证对象,对提出的控制策略与传统的统一变桨距控制策略进行仿真比较。结果表明,相对于统一变桨距控制,所提出的方法能有效确保在额定转速下机组输出功率稳定,同时也能有效降低风力发电机组各零部件的疲劳载荷。     

改进的遗传算法在渡槽结构优化设计中的应用

传统的遗传算法存在寻优效率较低、精度不高等缺点。为了解决这些问题,对传统遗传算法进行了改进——采用Fortran语言编制结构优化程序,并将改进后的遗传算法应用于大型预应力矩形渡槽优化设计中。通过对比原设计方案与优化方案的结构安全性及经济性,验证了优化设计方案具有显著的经济效益。结果表明,改进后的遗传算法提高了计算性能和优化效果。     

基于RBF多变量时间序列的滑坡位移预测研究

斜坡是一个受到多种因素影响的混沌动力系统,斜坡位移是其内部力学现象的宏观表现,具有很强的不确定性,从而导致难以建立斜坡位移的确定性方程。滑坡是斜坡的一种成因类型,具有相同的系统特性。滑坡经过防治后,其位移的主要外在动力因素除地下水外同时还受到防治设施的控制。滑坡位移及其影响因素所构成的混沌时间序列能够反映滑坡位移动力系统的历史行为。根据观测获得的多变量时间序列重构原滑坡位移动力系统,采用RBF神经网络实现变量间的映射关系,对滑坡位移进行了预测。预测结果对比分析表明:采用多变量时间序列预测模型能对滑坡位移进行有效预测,取得比单变量时间序列预测模型更好的预测效果;多变量时间序列预测模型具有更明确的物理力学意义,更能反映滑坡演化变形的实质特征。     

含分布式电源的配电网综合优化方法研究

讨论了不同位置和容量的分布式电源并网后对配电网的影响,对含有分布式电源的配电网络无功优化进行了研究。采用改进遗传算法,以综合经济效益最大为目标函数,将配电网中分布式电源选址定容与电容器投切相结合进行了综合优化。实例分析结果表明,该方法能够同时实现配电网络中分布式电源最优配置和配电网无功优化,从而有效降低配电系统损耗、改善电压质量,获得更好的综合经济效益。     

Asymmetric Network Interconnection

We develop a model of competition between interconnected networks,that allows for carriers to differ in size. Under two-partpricing, we show that because of asymmetry the larger network willalways prefer a reciprocal interconnection charge be set at cost.For sufficiently large asymmetry the smaller network will have thesame preference. Under the assumptions of our model a particularly simple regulation is optimal – if carriers cannot agree on the terms of interconnection, the larger carrier is entitled to select the access price which is then applied reciprocally.     

The Attitudes of British National Health Service Managers and Clinicians Towards the Introduction of Benchmarking

This paper describes an empirical study, conducted in three acute hospitals, of the attitudes of central managers, medical managers and clinicians towards the adoption of benchmarking. Benchmarking was portrayed in The New NHS White Paper (1997) as an important means of improving efficiency over the next decade. The present paper examines the context of change and nature of benchmarking. Findings are presented in seven sections including: the understanding which respondents had of benchmarking; their willingness to be involved in benchmarking; the existence of strategies and policies for implementing benchmarking; the relevance of existing costing information; and the role of networks in facilitating benchmarking. The study concludes that the process of change adopted contradicted most of the factors associated with creating receptivity to change. Also, that the publication of the National Reference Costs seemed to have more relevance to resource planning at central National Health Service Management Executive level, than to effecting improvements at operational level in acute hospitals.     

Informational Networks, Entrepreneurial Action and Performance

This paper develops a model that incorporates personal and business networks, firm action, and performance based on the existing literature. It explores the links between information and entrepreneurial-type action, and action and performance. Survey data was collected from a sample of 100 manufacturing firms in Thailand. Results show that entrepreneurs value the information they receive from their networks. However, there is little statistical support for tangible links between personal or business networks and entrepreneurial action and performance, or between action and performance.