改进的模糊C-均值聚类算法

为了克服模糊C-均值（FCM）聚类算法易陷入局部极小值和对初始值敏感的缺点,提出了一种基于改进量子蚁群的模糊聚类算法。将量子计算原理和蚁群算法相结合来改进FCM算法。初期采用量子遗传算法生成信息素分布,后期利用蚁群算法的全局搜索性、并行计算性等特点避免聚类陷入局部最优解。实验证明该算法保证了种群的多样性,有较好的全局收敛性,克服了模糊C-均值聚类算法的不足,能有效解决未成熟收敛的问题,使聚类问题最终快速、有效地收敛到全局最优解。     

自适应融合特征的人群计数网络

针对人群计数方法中存在的尺度变化和多层级特征融合不佳的问题,基于U-Net的编码器-解码器网络结构,提出一种自适应特征融合网络,来进行精准的人群计数。提出自适应特征融合模块,根据解码器分支的需要,高效地聚合编码器分支提取的高层语义信息和底层的边缘信息;提出自适应上下文信息提取器,从不同感受野下提取多尺度的上下文信息并自适应加权融合,提高网络对于人头尺度变化的鲁棒性。在ShanghaiTech、UCF-CC-50和UCG-QNRF上的实验表明,与目前主流的人群计数算法相比,该算法具有更强的准确性和鲁棒性。     

正交设计法优化从车前草中萃取熊果酸的工艺研究

用正交实验法优化采用超临界CO2技术萃取车前草中熊果酸的工艺条件。在单因素实验基础上,采取L9（3^3）探讨萃取压力、萃取温度、CO2流量对萃取车前草熊果酸的得率的影响情况,确定了超临界CO2萃取车前草中熊果酸的适宜条件。超临界萃取最优条件为：萃取压力35MPa,萃取温度为40℃,选择无水乙醇为夹带剂,夹带剂质量分数为20％,萃取时间2h,CO2流量为20L/h,产品的出膏率为3．58％,萃取物中熊果酸的含量为6．93％。该法与传统方法相比,具有操作简便快速,溶剂用量少,有效成分提取率高等优点。     

InfluenCe of Downstream Flow on Conduction Phase of Coaxial Plasma Opening Switch

Plasma behaviour and the scaling relations in a coaxial plasma opening switch （POS） using hydrogen plasma are studied self-consistently based on the two-dimensional magnetohydrodynamic （MHD） equations in conjunction with the generalized Ohm＇s law. The vacuum region on the right of POS is included in the model and the influence of downstream flow on the conduction characteristics is discussed. It is found that with the penetration of magnetic field, the pure hydrogen plasma is pushed downstream significantly; and POS still conducts current after the magnetic field arrives at the load edge of POS, which is different from the previous experimental results in a multispecies POS. It is because that the noticeable downstream plasma in the pure hydrogen POS may continue to conduct the current, while in the multispecies POS, the downstream plasma is unimportant so that the conduction phase ends soon after the magnetic field reaches the load edge. The scaling relation obtained from the simulations including the downstream region is consistent with the experimental results.     

Degradation of Anionic Dye Eosin by Glow Discharge Electrolysis Plasma

This paper describes a novel method for the degradation of eosin by using glow discharge electrolysis （GDE）. The effects of various parameters on the removal efficiency were studied. It was found that the eosin degradation could be raised considerably by increasing the applied voltage and the initial concentration, or by decreasing pH of the aqueous solution. Fe^2＋ ion had an evident accelerating effect on the eosin degradation. The degradation process of eosin obeyed a pseudo-first-order reaction. The relationship between the degradation rate constant k and the reaction temperature T could be expressed by Arrhenius equation with which the apparent activation energy Ea of 14.110 kJ· mol^-1 and the pre-exponential factor k0 of 2.065× 10^-1 min^-1 were obtained, too. The determination of hydroxyl radical was carried out by using N, N-dimethyl -p-nitrosoaniline （RNO） as a scavenger. The results showed that the hydroxyl radical plays an important role in the degradation process.     

Characteristics of Ions Emitted from Laser-Induced Silver Plasma

In this work, study of laser-induced ions is presented. The plasma was produced by focusing a Nd：YAG laser, with a wavelength of 1064 nm, a pulsed width of 9-14 ns, a power of 1.1 MW and energy of 10 mJ, on silver target in vacuum （10-3 Torr= 1.3332 Pa）. The characteristics of ion streams were investigated by CR-39 detectors located at angles of 0°, 30°, 60° and 90° with respect to normal of the target. The distance between the silver target and each detector was 11 cm. The energy of silver ions was found ranging from 1.5 eV to 1.06E4 eV. There was a high concentration of ions with low energy as compared to those with high energy, showing the energy distribution amongst the ions. The flux of ions was maximum in the axial direction which was decreasing with the angle increase with respect to normal of the target, and finally became minimum in the radial direction. Hence the silver ions have shown anisotropic behaviour.     

ABB-SYMPHONY系统在DEH改造中的应用

目前国内越来越多的电厂都在进行或即将进行自动控制系统的改造,而且自动控制系统的改造方案和技术也越来越成熟。针对巴蜀江油电厂330 MW机组DEH（ALSTOM公司产品）的结构和特点、改造实施以及ABB Bailey公司的DCS系统Symphony在巴蜀江油电厂DEH控制系统改造中的实际应用,进行技术交流,进一步了解进口机组控制系统的改造经验。     

基于灵敏度特征决策的GRNN短路容量智能辨识

针对当前局部地区短路容量水平已接近现有设备额定值的情况,提出一种短路容量智能辨识方法。利用基于潮流的短路计算法计算系统各母线的最大短路容量,通过对典型潮流下灵敏度的计算,选择对短路容量贡献程度较大的发电机、负荷的有功出力作为输入特征向量,建立训练样本,对广义回归神经网络（GRNN）进行训练,构成该电网结构下的短路容量辨识的人工神经网络。应用该模型对运行中电网的母线短路容量水平进行快速扫描,为智能电网与智能调度中的故障识别快速仿真建模（FSM）提供了一种新思路。通过IEEE 30节点系统验证了该方法的可行性与有效性。     

110 kV XLPE电缆载流量计算结果的验证试验

IEC 60287为电缆的载流量计算提供了标准方法,但由于计算所涉及的边界条件的可选性及实际运行中电缆周围环境的复杂性等原因,计算结果和电缆的实际承载能力是否一致,成为电缆行业所关心的问题。文章对影响载流量计算结果的因素进行了分析,并通过110 kV 1×400 mm2交联聚乙烯(XLPE)电缆进行了载流量计算结果的试验验证。结果表明,按IEC 60287标准计算所得的载流量完全可以作为线路设计、电缆选型的重要依据。     

基于免疫遗传算法的模糊C-均值聚类

为了克服FCM算法对初值的敏感性,提出了一种基于免疫遗传算法的FCM算法。该算法利用免疫系统原理和遗传算子自适应调整的方法（即免疫遗传算法）来改进FCM算法。实验证明该算法能有效解决未成熟收敛的问题,保证了种群的多样性,使聚类问题最终快速、有效地收敛到全局最优解。