基于格兰杰归因分析的高损台区窃电检测

配电台区中,用户窃电将增大台区线损电量,窃电电量直接对应非技术线损,并与台区线损电量间存在因果关系。分析指出,常见窃电手法中窃电用户的窃电量多与电表计量电量呈正相关关系,这将使得线损电量与窃电用户的计量电量之间存在隐含的因果关系。以此为基础,提出基于格兰杰归因分析的高损台区窃电检测方法。通过协整检验及格兰杰检验分析用户用电量与台区线损时间序列间的长期趋势及格兰杰因果关系,将对台区线损影响突出的用户识别为窃电用户。基于实际高损台区和高损线路用户数据的测试算例和现场稽查验证了所提方法的有效性。     

基于边缘计算的低压用户窃电检测

低压用户窃电导致线损电量增加,对台区线损异动进行归因分析是识别窃电用户的有效途径。低压用户通信异常多发,可导致用电信息采集系统主站数据失真,易误导窃电检测。利用配变终端可就地完整准确采集台区数据的特点,提出基于边缘计算的低压用户窃电检测方法。首先,在通信正常和异常的条件下,分析台区窃电用户用电量与线损电量的关联关系;然后,在配变终端窃电检测模块中对真实的台区线损和用户用电量进行归因分析来识别窃电用户;最后,基于高损台区实际数据的仿真分析,验证了所提方法相比于在主站侧采用异常数据以及采用不同缺失数据填补算法修复后的数据进行窃电检测时的优势。     

基于向量自回归模型的高损线路窃电检测

接入配电线路的用户窃电将直接导致分线线损电量异常波动,因此用户用电量与分线线损电量间存在长期动态互动。该文提出基于向量自回归模型检测造成线损波动的异常用户的方法。首先,运用边限协整检验分析线损电量和用户用电量的长期均衡关系;然后,构建线路线损电量与接入用户用电量的向量自回归模型,计算脉冲响应函数分析线损电量与用户用电量的动态作用机制;再通过方差分解分析线损电量与用户用电量间冲击作用的贡献度,将对线损电量有显著影响且波动贡献度最大的用户识别为窃电用户;最后,结合实际高损线路用户数据识别窃电嫌疑用户,并运用现场稽查的手段验证了所提方法的有效性。     

基于局部离群点检测的低压台区用户窃电识别

针对低压台区反窃电工作效率低、成本高的问题,提出了一种基于局部离群点检测的低压台区用户窃电识别方法。首先,基于线损特征搜寻与待检测台区最相似的k个最近邻台区;接着,基于k个最近邻台区的线损率分析待检测台区某段时间线损率是否异常,若线损率异常计算该台区所有用户该段时间负荷曲线之间的离散Fréchet距离;最后,基于离散Fréchet距离计算台区每个用户负荷曲线的局部离群点因子,同时采用邻域查询优化技术提高计算效率,局部离群点因子越大的用户窃电嫌疑越大。该方法输出待检测台区所有用户窃电嫌疑度排序,只需要检测排序靠前的用户即可检测出大部分窃电用户,大大提高了反窃电工作效率。     

基于皮尔逊相关系数法的台区窃电精准分析

正针对台区损失电量与用户用电量、台区线损率与用户用电量之间的变化关系,利用皮尔逊相关性系数进行量化,并基于Excel利用VBA语言编写相关程序,实现高损台区异常用户的一键分析,提高科学反窃成效。皮尔逊相关性系数(Pearson)判断原理:Pearson其值在(-1,1)之间;Pearson越接近于1,代表呈正相关,即该用户窃电或因计量异常少计电量的嫌疑程度越高;Pearson越接近于-1,代表呈负相关,即该用户间断性窃电或户变关系异常的嫌疑程度越高;Pearson越接近于0,     

基于SE-CNN模型的窃电检测方法研究

针对传统电网现有窃电检测仅用单一电量且实际数据集下检测准确度低的问题,提出一种基于通道注意力网络改进卷积神经网络模型的窃电行为检测方法。首先建立一种包含用电量趋势、线损增长率、终端告警多源数据融合的窃电评价指标体系,以此构建用户用电特征集。然后,基于通道注意力挤压激励网络(squeeze and excitation networks, SENet)优化卷积神经网络(convolutional neural network, CNN)模型,据此构建自适应通道注意力网络改进卷积神经网络模型的窃电检测方法。最后,利用南方电网数据集对提出方法的有效性与准确性进行验证。实测结果表明,所提方法能有效实现实际电网情况下各类窃电行为准确检测,建立的评价指标体系可更明显表征窃电行为规律。构建的窃电检测模型可自适应对特征通道重要程度调整,提升通道利用率。相比现有检测方法,其具有更高准确度和更优泛化性能。     

基于堆叠稀疏自编码器和深度森林的窃电检测模型

为了提高窃电检测模型的检出率和准确率，提出一种基于堆叠稀疏自编码器和深度森林的窃电检测模型。通过使用用户用电量数据训练堆叠稀疏自编码器，以降低总体重建误差为目标，确定堆叠稀疏自编码器的层数及隐藏层输出向量维度，实现对用户用电特征更有效的提取。将最后一层稀疏编码器隐藏层的输出向量作为深度森林的输入特征向量，对输入特征向量进行多粒度扫描，使用级联森林实现特征深度融合与窃电检测。基于某市6 000个用户用电量数据对该模型的性能进行验证，实验结果表明，对比其他经典窃电检测模型，所提窃电检测模型具有更高的检出率和准确率。     

基于负荷尖峰特征LSTM自编码器的窃电识别方法

近年来，面向高损线路的窃电检测方法得到大面积工程应用，对降低窃电检测误报率和推动数据驱动窃电检测的工程应用起到了重要作用。但如何准确检出非高损线路的专变窃电用户，仍是亟待解决的难题。基于实践经验中部分窃电用户存在用电量异常尖峰这一特点，提出基于负荷尖峰特征长短期记忆（LSTM）自编码器的用户窃电识别方法。首先，分析典型窃电用户曲线形态,提炼了区分正常及窃电用户的用电量尖峰特征。然后，结合该特征和用户分时数据周期性规律，构建LSTM自编码模型重构输入得到拟合值，基于拟合值与真实值的均方误差设定自适应阈值，从而识别窃电嫌疑用户并提供具体预警尖峰时段。最后，应用实际专变用户用电数据进行算例分析，结果表明所提方法在准确率、命中率和误报率上均优于对比方法。     

基于离群点算法和用电信息采集系统的反窃电研究

为了解决以往依靠用户每月电量变化、台区线损异常确定窃电嫌疑户等传统方法时效性差、准确率低,现场排查发现窃电行为又效率不高的问题,利用当前远程集抄海量数据为传统反窃电理论研究和高效反窃电的实现尝试探索一种新方法。全面分析窃电手段及原理,绘制实施窃电手段的鱼骨图,由实际案例统计分析确定研究重点为欠压、欠流窃电法。根据高供高计和高供低计用户计量装置接线特点、电压电流值的规律以及离群点理论确定窃电判定算法,并做出基于距离的离群点检测法判定窃电的流程图。通过实例验证提出的算法和窃电户筛选流程能够完全甄别出窃电用户,为监控人员利用用电信息采集系统的海量数据精确及时地进行有效反窃电分析提供了新思路。     

基于线损协方差分析的群体性固定比例窃电行为检测方法

当一个配电台区内存在多个固定比例的窃电用户时,台区的非技术线性损耗（NTL）将由这群窃电用户的窃电量共同决定。目前根据NTL相关性强弱排序的检测方法在这一场景下有可能失效。首先,为了解决这一问题,文中发现了台区NTL和固定比例窃电用户表计数据的相关性存在着一种递增现象,并论证了这一现象成立的充分条件。以此为基础提出了基于协方差分析的窃电检测方法。通过标准化处理后的协方差来衡量NTL与用户电量之间的相关性,以求解一个组合优化问题的方式,实现对固定比例窃电用户的搜寻。然后,设计了该方法在实际应用过程中的运行框架和检测方案。最后,基于中国某省电力公司提供的实测数据和窃电模拟实验平台生成的窃电数据验证了所提方法的有效性。     

On a wing and a prayer

The article talks about recent Defense Research Projects Agency (DARPA) presolicitation notice in the area of hybrid insect micro-electromechanical systems (MEMs). DARPA's goal is to fund promising interdisciplinary research that will permit the creation of "insect-cyborgs"     

利用光伏电池给蓄电池充电

磷酸锂铁（LiFePO4）电池对比钻基德Li离子/聚合物电池,具有安全性性高、浮充电压小、自放电率低、使用寿命长等一系列优点。本文所述LT36521C为可实现太阳能跟踪、单块式补偿蓄电池充电的集成电路。当LT3652由单块太阳能电池板供电时,其输入调节回路能确保在最大输出功率下工作,充分体现出这一新型太阳能充电器的优越特性。     

Making a Case for a New Connector

The emergence of quick lock connectors, particularly quick lock versions of the SMA and N connectors (QMA and QN, respectively), was an inevitable development in the RF connector industry and is of significant importance in that nearly 50% of all practical applications for RF connectors are currently supported by either SMA or N. Once QMA and QN can achieve similarity in cost and performance to the present SMA and N connectors, they will take the place of their predecessors in most applications, creating a huge new market. Our RF team has long been dedicated to research and development of these two connector types in the hopes of creating an optimal design. In this article, we will show some of the results of our research.     

Chaos in a series circuit with a nonlinear capacitor and a nonlinear inductor

This paper proposes a nonlinear circuit that generates chaotic oscillations. The proposed nonlinear circuit includes a nonlinear capacitor and a nonlinear inductor by using Generalized Impedance Converters. Both computational experiments and experiments on real circuits show that the region for chaotic oscillation is very wide, demonstrating that the proposed circuit is effective as a chaos generator. © 2004 Wiley Periodicals, Inc. Electr Eng Jpn, 150(2): 35–42, 2005; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/eej.10361     

Electric potential near a tip of a local discharge on a conductive film

To develop a propagation model of a local discharge on a wet polluted insulator, the authors have examined characteristics of the local discharge on an electrolytic solution simulating the wet polluted insulator surface. The local discharge was propagated on the solution under the impulse voltage. A shape of a contacting surface between the local discharge and the solution has been already obtained by the authors. By using the contacting surface, an electric potential near a tip of the local discharge on a conductive film of 1 mm in thickness was calculated. Furthermore, the potential was measured with a probe. The calculated values and the measured values of the electric potential agreed well regardless of the propagation length of the local discharge, the width of the conductive film, the resistivity of the conductive film, the peak value of the applied voltage and the applied voltage waveform.     

Power losses in a conducting ring enclosing a conductor with a transient current

Contents In the paper Joule power losses are calculated in a conducting ring enclosing a conductor with a transient current having an alternating component.—Theoretical calculation and numerical computation are worked out basing on a link Bubnov-Galerkin method in its numerical version using the finite elements for the conductor with the Time-Stepping algorithm for the time discretization.—On the basis of numerical computation, graphs of the relative power losses are plotted as a time function.

---

Leistungsverluste in einem leitenden Ring, der einen Leiter mit einem instationären Strom umschließt

Übersicht In diesem Beitrag werden die Joulschen Leistungsverluste in einem leitenden Ring berechnet, der einen Leiter umschließt, in dem ein instationärer Strom mit Wechselanteil fließt. Die theoretischen Betrachtungen und numerischen Berechnungen hat man in Anlehnung an die Bubnov-Galerkin-Methode (in ihrer numerischen Variante, die von einen Leiter unter Verwendung der Methode der finiten Elemente realisiert wird) in Verbindung mit einem Schritt-Algorithmus mit der diskretisierten Zeit ausgeführt. Auf der Grundlage numerischer Berechnungen sind Diagramme für die relative Leistung in Abhängigkeit von der Zeit dargestellt.

---

List of symbols Symbol Unit Physical parameter - B Wb/m² magnetic induction vector - H A/m magnetic field intensity vector - H, H A/m -component of the magnetic field vector - E V/m electrical field vector - E _r ,E _z V/m components of the electric field - f s^–1 frequency - I A current intensity (r.m.s. value) - L H inductance - P W power losses - P ₀ W power losses for d.c. - R resistance - t sec time - ^–1 m_–1 electric conductivity of the conductor - Hm^–1 magnetic permeability - _ijk m² finite element area - basis function - rad s^–1 pulsation - N _i ,N _j ,N _k function of finite element shape - 3.141593... - region - ² Laplace's operator - r, ,z cylindrical coordinates - ^h region approximating the region - ^e finite element region - M number of discretization points of the region - V number of finite elements - (N–1) number of steps - i, j, k indices of vertices of triangular finite element     

Development of a deceleration control system for a winding machine based on a microcontroller

The development of an automatic deceleration system for a winding machine in a siderurgy factory is presented in this paper. The proposed system controls the speed of a winding machine in order to obtain the optimum stop point in terms of security and fast production. The system is based on a microcontroller and an extremely simple hardware. The system has been tested for two years and results evince a very good mean time between failures.     

Simulation of a generator for a wind-power unit

Analysis of excitation systems of generators for wind-power units is carried out, a software package for generator simulation is presented, and the sequence of the generator design is presented.     

When a feature detector becomes a feature generator

The neural network approach is reviewed in the context of studies in the animal and human visual systems. An attempt is made to explain perception, memory, and learning as a feedback loop similar to the one used in the ALOPEX optimization concept. Applications in pattern recognition and image processing are discussed. Experiments using visual evoked potentials in humans are described.