适用于发展性故障的故障分量提取算法

提出了适用于发展性故障的故障分量提取算法。该算法首先计算发展性故障发生前保护安装处的故障分量,再根据发展性故障发生前保护安装处的电流和获取的故障分量求得非故障状态网络中的支路电流;发生发展性故障后,利用保护安装处的全电流减去求得的支路电流得到发展性故障后的故障分量电流。该算法不受电网参数波动和负荷变化的影响。在IEEE 9节点系统模型基础上,BPA仿真验证了算法的正确性。     

基于工频量补偿算法的长线距离保护

特高压长线路的分布参数特性使传统距离保护的测量阻抗不与故障距离成正比,容易造成保护的超越或拒动.文中分析距离保护正确动作的条件:测距误差不大于故障距离与整定距离之差,提出将保护安装处的电流、电压补偿到距离Ⅰ段末端测距.仿真表明,该算法提高了距离Ⅰ段末端的测距精度,符合保护正确动作的条件,使保护正确动作.     

图像分类模型的对抗样本攻防研究综述

深度学习模型在图像分类领域的能力已经超越了人类,但不幸的是,研究发现深度学习模型在对抗样本面前非常脆弱,这给它在安全敏感的系统中的应用带来了巨大挑战。图像分类领域对抗样本的研究工作被梳理和总结,以期为进一步地研究该领域建立基本的知识体系,介绍了对抗样本的形式化定义和相关术语,介绍了对抗样本的攻击和防御方法,特别是新兴的可验证鲁棒性的防御,并且讨论了对抗样本存在可能的原因。为了强调在现实世界中对抗攻击的可能性,回顾了相关的工作。在梳理和总结文献的基础上,分析了对抗样本的总体发展趋势和存在的挑战以及未来的研究展望。     

Effects of ethyl acetate additive on Cu2ZnSnS4 solar cells fabricated with a facile dimethylformamide-based solution coating process

The solution based on dimethylformamide (DMF) has shown promising application prospects in the fabrication of high-efficiency Cu₂ZnSnS₄ solar cells. However, due to the high boiling temperature of the solvent, it is difficult to completely volatilize DMF during the evaporation process after spin coating, leading to remains of C and O atoms at grain boundaries, which severely restricts the photoelectric performance of solar cells. In this study, ethyl acetate (EA) with green character was used as an additive to optimize the film formation process of DMF-based CZTS precursor. The experimental results showed that using a small amount of EA additives could effectively improve morphology, crystallinity, composition distribution and electrical properties of the CZTS absorber. In addition, the CZTS and CdS heterojunctions exhibited a cliff-like energy band structure, and the optimized conduction band offset increased the activation energy required for the carrier recombination path, consequently reducing the carrier recombination. Compared with the pure DMF precursor solution, the photoelectric conversion efficiency of CZTS solar cells with an EA addition ratio of 10% was improved by 42%, and the open circuit voltage of the device reached 601 mV.     

一种结合空间与光谱信息的改进CVA变化检测方法

基于变化向量分析（CVA）的变化检测方法通过直接比较像素差异，能够快速提取多时相影像间的变化信息。尽管如此，由于忽略了像素领域的空间上下文信息及波段之间的差异性和互补性，导致检测结果中难以消除噪声等因素产生的“伪变化”。为此提出了一种结合空间和光谱信息的改进CVA方法。首先，采用主成分分析法对影像进行增强，继而通过构建一种新的多方向差分描述子来提取中心像素的空间上下文信息；在此基础上，提出一种基于相关性的加权融合策略，获得统一的变化强度差分影像；最后，采用EM算法求得变化像素的阈值，继而得到二值检测结果。实验结果表明：所提出的算法能够有效应对“伪变化”的干扰，显著提高变化检测的精度及可靠性。     

DS-CDMA通信系统中基于独立分量分析的盲多用户检测

本文研究了直接序列扩频码分多址(DS-CDMA)通信系统的多用户检测问题,提出了基于独立分量分析(ICA)方法的盲多用户检测算法,实现多用户信号的盲检测。该方法不需知道用户的扩频码就可完成多用户信号的盲检测。仿真结果验证了本文提出的方法的优良性能。     

Fault component integrated impedance-based pilot protection scheme for EHV/UHV transmission line with thyristor controlled series capacitor(TCSC) and controllable shunt reactor(CSR)

This paper analyzes the fundamental frequency impedance characteristic of a thyristor controlled series capacitor(TCSC) and presents a novel transmission line pilot protection scheme based on fault component integrated impedance(FCII) calculated for a transmission line with TCSC and controllable shunt reactor(CSR).The FCII is defined as the ratio of the sum of the fault component voltage phasors of a transmission line with TCSC and CSR to the sum of the fault component current phasors where all the phasors are determined at both line’s terminals.It can be used to distinguish internal faults occurring on the line from external ones.If the fault is an external one the FCII reflects the line’s capacitive impedance and has large value.If the fault is an internal one on the line the FCII reflects the impedance of the equivalent system and the line and is relatively small.The new pilot protection scheme can be easily set and has the fault phase selection ability and also it is not affected by the capacitive current and the fault transition resistance.It is not sensitive to compensation level and dynamics of TCSC and CSR.The effectiveness of the new scheme is validated against data obtained in ATP simulations and Northwest China 750 kV Project.     

一种分块差分滤波器及其在仅有角度跟踪中的应用

为解决仅有角度跟踪时,目标估计受限于较大的初始估计误差和噪声统计特性未知的问题,提出了一种带有噪声智能统计功能的改进型分块差分滤波器.通过统计线性化方法得到了一种S-H智能噪声统计估值器,并用其优化传统分块噪声滤波器的测量更新步骤,实现了对未知过程和测量噪声的智能统计处理,通过迭代更新进一步提高了滤波器对于复杂非线性函数的适应能力.与目前几类主流的自适应滤波器性能相比,结果表明:对于具有线性系统模型和非线性测量模型的典型被动跟踪估计问题,针对较大的初始状态估计误差,所给出的滤波器能更好地完成系统噪声和测量噪声部分参数统计特性未知情况下的非线性估计任务,在保证计算量适中的同时有效地提高跟踪制导精度.     

Evaluating the Role of Nanostructured Current Collectors in Energy Storage Capability of Supercapacitor Electrodes with Thick Electroactive Materials Layers

Electroactive materials (especially pseudocapacitive materials) are generally in the form of ultrathin conformal coating in supercapacitor electrodes based on nanostructured current collectors; thus, the resultant low mass loading of electroactive materials largely limits the applications of nanostructured current collectors. Here, supercapacitor electrodes with nickel nanorod arrays as nanostructured current collectors and MnO₂ as electroactive materials are fabricated to study the role of nanostructured current collectors in determining the energy storage capability when electroactive materials are in thick layer rather than ultrathin conformal coating. Electrochemical analysis reveals that Ni nanorods could create numerous electrical conductive tunnels in the thick‐layer electrodes to dramatically alleviate the contact resistance at the electroactive‐material/current‐collector interface. With 1 µm thick MnO₂ layer, the Ni nanorods based electrodes have much higher areal capacitance than those with Ni foils as current collectors, which is more than six times of that with the same MnO₂ mass loading or more than 18 times of that with the same 1 µm thick MnO₂ layer. Moreover, better rate capability and higher structural stability is maintained in Ni nanorods based electrodes even with 3 µm thick MnO₂ layer. These results open up new opportunities for nanostructured current collectors to construct supercapacitors with superior energy storage capability.     

一种基于Spark/DBSCAN算法的城市公交扒窃行为分析方法

随着中国城镇化进程的不断加速,城市公交扒窃行为给公民的生命财产安全构成极大的威胁,对社会风气造成了持续性的影响.研究中基于"犯罪行为是行为人主观意识对客观环境的反致行为"[1]这一理论,采用NLPIR汉语分词系统,整理以往城市公交扒窃行为数据,提取犯罪行为时间、位置数据,在Spark分布式计算框架下,利用DBSCAN算法做地理位置数据的聚类分析,发现造成犯罪反致行为[1]的客观环境和规律.通过改善客观环境,进而提高预防和控制犯罪行为的水平.