用户电力技术及其在电力市场环境中的应用

根据电力市场中的电能质量的问题的特点，阐明了用户电力技术是解决电力市场环境中电能质量问题的一个很好的途径；提出了在不同的电力市场运营模式下用户电力技术的实现方式，以及需要遵循的原则。     

统一电能质量控制器的建模与仿真

统一电能质量控制器可同时补偿电网畸变电压和抑制负载谐波电流。为此,构造了一种基于反向传播算法的三层前馈神经网络用来检测并联型有源电力滤波器的谐波电流,离线训练收敛后实现在线功能,对串联型有源电力滤波器谐波电压检测采用了畸变电压参考量比较检测方法;建立了统一电能质量控制器的系统仿真模型,利用其对各种电能质量问题的补偿性能进行了仿真研究,并对补偿前后负载和电源电流/电压进行了频谱分析。研究结果表明,统一电能质量控制器集电压补偿、电流补偿于一体,可有效实现多重电能质量调节功能。     

模块化PWM主电路实现的大容量有源电力滤波器

有源电力滤波器在抑制电网谐波，提高电能质量方面有良好的应用前景。提高有源电力滤波器的装置容量是应用中有待解决的关键技术问题之一。文中在综合国内外目前实现大容量有源电力滤波器方法的基础上，介绍了一种采用多个模块化PWM主电路单元组合实现的大容量并联型有源电力滤波器，给出了该有源电力滤波器的主电路结构拓扑，分析了系统的工作原理及控制方法，并对在实现该装置时所遇到的问题进行了讨论。工业现场实验结果表明，采用该方法实现的有源电力滤波器有着良好的补偿特性，有效地解决了有源电力滤波器在大容量时器件与装置容量之间的矛盾。     

基于DSP的多功能电能表的设计与实现

论述了多功能电能表在谐波工况下的电能计量方法,重点分析了基于Go-ertzel算法的电力参数和谐波参数的计算方法,讨论了基于DSP技术的多功能电能表电能计量模块的设计方案.最后给出了硬件和软件的实现方法.     

基于模糊理论及神经网络的电力短期负荷预测方法

分析了电力短期负荷的构成及其影响因素的特点，结合模糊技术和神经网络。设计了用于电力短期负荷预测的神经网络。利用模糊技术及相似性原理对作为BP神经网络训练样本的电力负荷按天气特征进行分类识别，同时对带有噪音的训练样本提出了处理方法。针对BP神经网络的不足，提出了改进算法；并用实际例子对上述电力短期负荷预测神经网络进行了例证。     

输出电压恒定的电力电子变压器仿真

介绍了电力电子变压器的优点、目前的研究状况；指出了用电力电子变压器解决电能质量问题是今后的发展趋势；着重讨论了电力电子变压器的工作原理，推导了高频变压器中的磁链关系式，比较了采用基于有效值和瞬时控制的电力电子变压器的动态响应，并用仿真验证了理论推导的正确性和所采用的控制策略的有效性。结果表明，电力电子变压器体积小、重量轻，不但可以实现电能传输、电压隔离、匹配等功能，还能抑制电压跌落、上升和闪变等动态电能质量对输出电压的影响，是建设“绿色电网”、“数字电网”的关键设备之一，对其进行研制和使用可以取得巨大的经济和社会效益。     

基于神经网络电流检测的有源电力滤波器研究

给出了一种用于有源电力滤波器的神经网络谐波及无功电流检测方法，并对基于该方法的并联型有源电力滤波器进行了研究。给出了仿真结果和实验结果。     

基于Profibus-DP的电能管理与电力监控系统

设计了基于Profibus-DP的电能管理及电力监控系统。描述了该系统的结构组成和实现原理,给出了主站和串口电力仪表从站通信的实现方法,解决了主从站通信程序设计中的关键问题。经试验验证了系统的通信性能和可行性。     

融合神经网络与电力领域知识的智能客服对话系统研究

为了开发面向电力领域的智能客服对话系统,提出了融合神经网络和电力领域知识的方法。利用基于神经网络的自然语言理解技术从客户的话中提取有用信息;利用对话流配置框架和知识图谱将电力客服业务流程、客服数据库和标准问答对等电力领域知识融入对话系统,指导对话系统给出正确的回复。最后通过实现两个电力领域客服对话系统验证了方法的可行性。     

基于小波和神经网络的电能质量扰动信号数据压缩

在小波变换数据压缩方法和神经网络数据压缩技术的基础上,提出了将小波和神经网络应用于电能质量扰动信号数据压缩的方法。利用小波时域和频域的双重分辨率和神经网络的非线性函数逼近能力,以压缩比、均方误差为压缩效果的评价指标,对实际扰动信号进行数据压缩。采用样条小波和径向基神经网络数据压缩方法,以一个实例,给出了电能质量扰动信号的压缩仿真过程,给出了各类(电压凹陷、突起、尖峰、闪变及瞬态振荡)电能质量扰动信号的仿真分析结果。结果表明,该电能质量扰动信号数据压缩方法,压缩后得到的均方误差为-16.1397 dB,压缩效果良好。     

火电厂烟气含氧量测量的现状与发展

燃煤电厂烟气含氧量的准确测量对实现燃烧系统的闭环控制和优化运行具有重要的意义。分析了烟气含氧量测量方法的现状,传统的氧量分析仪存在一些问题,给出了软测量建模技术的一般理论和方法,讨论了烟气含氧量的软测量方法,包括机理分析、统计回归分析、神经网络、支持向量机等建模方法。由于单一测量方法均存在缺点,两种以上测量方法相混合的烟气含氧量技术已被使用,它也是烟气含氧量测量的发展趋势。     

电网信号四象限功率测量原理及实现方法

针对传统电力仪表测量的功率不能正确反映电能计量点真实功率状态的缺陷,提出一种用于电网信号功率测量与分析的四象限功率测量原理及实现方法.首先深入分析了四象限功率测量原理.四象限功率测量原理从矢量的角度对功率进行分析,惟一确定了电能交换的真实状态和数值,提高了功率测量的准确度和适用性,是测量与分析电网信号功率的一种有效方法,既适用于功率传输方向确定的电能计量点的功率测量,也适应于功率传输方向变化的电能计量点的功率测量.然后给出了2种实现方法:间接法和直接法.最后给出了仿真验证结果,表明了四象限功率测量原理及实现方法的正确性和可行性.     

基于集成深度神经网络的配电网分布式状态估计方法

随着大量分布式能源的接入,配电系统的运行与控制方式愈加复杂。针对配电网状态估计方法面临分布式电源波动数据辨识困难、估计精度低、鲁棒性与估计时效性差等问题,提出一种基于集成深度神经网络的配电网分布式状态估计方法。首先,利用量测数据相关性检验的数据辨识技术识别不良数据和新能源波动数据。在此基础上,利用时域卷积网络(temporal convolutional network, TCN)-双向长短期记忆网络(bidirectional long short term memory, BILSTM)对不良数据进行修正。然后,建立集成深度神经网络(deep neural network, DNN)状态估计模型,采用最大相关-最小冗余(maximum relevance-minimum redundancy, MRMR)的方法优化训练样本,从而提高状态估计的精度和鲁棒性。最后,建立分布式集成深度神经网络模型,弥补了集中式状态估计速度慢的不足,从而提高状态估计效率。基于IEEE123配电网的算例分析表明,所提方法能更准确地辨识分布式电源波动数据和不良数据,同时提高状态估计的精度和效率,且具有较高的鲁...     

Artificial neural networks for solving the power flow problem in electric power systems

In this paper, the use of artificial neural networks (ANN) is proposed for solving the well known power flow (PF) problem of electric power systems (EPS). PF evaluates the steady state of EPS and is a fundamental tool for planning, operation and control of modern power systems. The mathematical model of the PF comprises a set of non-linear algebraic equations conventionally solved with the Newton-Raphson method or its decoupled versions. In order to take advantage of the superior speed of ANN over conventional PF methods, multilayer perceptrons neural networks trained with the second order Levenberg–Marquardt method have been used for computing voltages magnitudes and angles of the PF problem. The proposed ANN methodology has been successfully tested using the IEEE-30 bus system.     

基于小波神经网络的电力电子电路故障模式识别

提出了基于两种不同小波神经网络的电力电子电路故障模式识别方法。针对电力电子电路故障,构造了激活函数型和权值型两种不同的三层小波神经网络,给出了相应的数学模型和学习算法。以三相整流桥电路为例,建立了小波神经网络的输出与故障元之间的对应关系,实现了电路故障的模式识别,并与用普通BP网络识别的结果进行了比较。仿真结果验证了两种故障识别方法的正确性和较好的准确性。     

Fault location observability analysis and optimal meter placement based on voltage measurements

Various transmission line fault location algorithms have been proposed in the past depending on measurements available. These methods evince that if a sufficient number of meters are placed in a power network to record the fault measurements, then the fault location can be reliably estimated. A relevant question to ask may be: how many meters are sufficient in order to derive a reliable and unique fault location estimate for a given network? This paper addresses this question by defining and performing the fault location observability analysis. An optimal meter placement scheme is proposed for determining the optimal locations to place meters so as to make the system observable while minimizing the required number of meters to reduce costs. The proposed method is especially useful for power networks where digital relays have not yet been widely adopted and measuring devices such as digital fault recorders are deployed for monitoring purposes. A sample power network has been employed to illustrate the proposed method.     

基于同伦神经网络的短期负荷预测研究

将同伦方法引入到神经网络的学习训练中,提高了神经网络的学习效率,较好地解决了BP神经网络收敛速度慢和局部极小值等问题.同伦神经网络具有稳定性强,收敛性好的特点,用这种神经网络进行了电力系统短期负荷预测,算例表明了网络收敛快、预测精度高的优越性.     

发电机组启停机的智能优化经济调度研究

遗传算法用于解决启停机优化调度问题中,常出现群体早熟和有时收敛于局部最优解等问题.而模拟退火算法在接受新解时却显示出较好的特性.在遗传算法的评价函数中引入模拟退火算法以及在选择操作中采用模拟退火算法的接受准则,将两者进行混合可有效地缓解其选择压力,增强算法的全局收敛性.采用十进制编码,无需解码,可减少计算误差和时间.算例分析表明该算法可以在满足安全可靠的多种约束条件下,较好地改善机组启停计划的经济性.     

One-hour-ahead load forecasting using neural network

Load forecasting has always been the essential part of an efficient power system planning and operation. Several electric power companies are now forecasting load power based on conventional methods. However, since the relationship between load power and factors influencing load power is nonlinear, it is difficult to identify its nonlinearity by using conventional methods. Most of papers deal with 24-hour-ahead load forecasting or next day peak load forecasting. These methods forecast the demand power by using forecasted temperature as forecast information. But, when the temperature curves changes rapidly on the forecast day, load power changes greatly and forecast error would going to increase. In conventional methods neural networks uses all similar day's data to learn the trend of similarity. However, learning of all similar day's data is very complex, and it does not suit learning of neural network. Therefore, it is necessary to reduce the neural network structure and learning time. To overcome these problems, we propose a one-hour-ahead load forecasting method using the correction of similar day data. In the proposed prediction method, the forecasted load power is obtained by adding a correction to the selected similar day data     

PREDICT – Decision support system for load forecasting and inference: A new undertaking for Brazilian power suppliers

One of the most desired aspects for power suppliers is the acquisition/sale of energy for a future demand. However, power consumption forecast is characterized not only by the variables of the power system itself, but also related to social–economic and climatic factors. Hence, it is imperative for the power suppliers to project and correlate these parameters. This paper presents a study of power load forecast for power suppliers, considering the applicability of wavelets, time series analysis methods and artificial neural networks, for both mid and long term forecasts. Both the periods of forecast are of major importance for power suppliers to define the future power consumption of a given region. The paper also studies the establishment of correlations among the variables using Bayesian networks. The results obtained are much more effective when compared to those projected by the power suppliers based on specialist information. The research discussed here is implemented on a decision support system, contributing to the decision making for acquisition/sale of energy at a future demand; also providing them with new ways for inference and analyses with the correlation model presented here.