Comparison of modeling techniques in circuit variability analysis

Three nonlinear reduced‐order modeling approaches are compared in a case study of circuit variability analysis for deep submicron complementary metal‐oxide‐semiconductor technologies where variability of the electrical characteristics of a transistor can be significantly detrimental to circuit performance. The drain currents of 65 nm N‐type metal‐oxide‐semiconductor and P‐type metal‐oxide‐semiconductor transistors are modeled in terms of a few process parameters, terminal voltages, and temperature using Kriging‐based surrogate models, neural network‐based models, and support vector machine‐based models. The models are analyzed with respect to their accuracy, establishment time, size, and evaluation time. It is shown that Kriging‐based surrogate models and neural network‐based models can be generated with sufficient accuracy that they can be used in circuit variability analysis. Numerical experiments demonstrate that for smaller circuits, Kriging‐based surrogate modeling yields results faster than the neural network‐based models for the same accuracy whereas for larger circuits, neural network‐based models are preferred as, in all metrics, better performance is obtained. Within‐die variations for an XOR circuit are analyzed, and it is shown that the nonlinear reduced‐order models developed can more effectively capture the within‐die variations than the traditional process corner analysis. Copyright © 2011 John Wiley & Sons, Ltd.     

基于混沌优化BP神经网络的江河涌潮短期预报模型

为提高江潮潮时预报的准确性,针对经验模型和传统神经网络模型预测精度较差的局限性,本文采用基于相空间重构技术的BP神经网络模型预报江潮潮时。该模型首先对隔日的到潮时差序列进行混沌特性分析,然后利用重构相空间来确定BP神经网络的输入结构。该模型给出了到潮时差序列可能的误差预测,修正最终预报结果。通过对钱塘江四个观测站潮时预测,四个站点潮时统计的RMSE值平均减少了83.9%,表明该模型可靠且具有较高的预报精度。     

基于残差BP神经网络的变压器故障诊断

基于传统BP神经网络的变压器故障诊断方法,当网络模型达到一定的深度时,模型的诊断性能会趋向于饱和,无法进一步提升网络模型的诊断性能,此时加深网络模型的深度反而会导致模型的诊断性能有所下降。此外,在小样本数据下,传统BP神经网络仍无法取得较好的诊断准确率。因此,为了提高变压器故障诊断准确率以及在小样本数据下的诊断性能,提出了基于残差BP神经网络的变压器故障诊断方法。所提方法采用堆叠多个残差网络模块的方式加深BP神经网络的深度,将传统BP神经网络的恒等映射学习转化为残差BP神经网络中的残差学习。同时,在每个残差网络模块中,模块的输入信息可以在模块内跨层传输,使得每个模块的输入信息可以更好地向深层网络传递,从而在小样本数据下仍可以训练得到较好的诊断模型。实验结果表明,相较于传统深层BP神经网络和传统浅层BP神经网络,所提方法具有更高的诊断准确率,同时在小样本数据下也体现出较好的诊断性能。     

基于Bayes-LSTM网络的风电出力预测方法

为提高风电出力的预测精度,提出一种基于Bayes优化的长短期记忆人工神经网络(long-short term memory, LSTM)的预测模型。首先,利用经验模态分解对风电历史出力序列进行分解,并对各分量及原始数据分别提取8个统计特征量,与预测前6个时刻出力值共同组成预测特征集。然后,采用绳索算法(least absolute shrinkage and selection operator, LASSO)从预测特征集中提取具有统计意义的特征子集,作为预测模型的输入。最后,提出基于Bayes超参数寻优的LSTM网络优化方法,以提高预测精度。选取湖北某市风电出力历史数据进行预测实验,结果表明：相较于BP神经网络、SVM、RBF网络、GRNN网络等预测模型,所提模型预测精度较高,特征提取方法较为合理。     

基于掩模经验模态分解的风速组合预测模型

在风电预测中,传统的经验模态分解法将风速信号分解为若干具有不同特征尺度的数据分量时,其所得分量可能存在模态混叠现象,影响风速预测的精度。为此,提出一种基于掩模经验模态分解法和遗传神经网络的风速预测组合模型。首先,通过掩膜信号法（masking signal,MS）对经验模态分解法进行改进,将风速信号分解为频率相对固定、更为平稳的分量。之后,利用遗传神经网络算法分别对这些分量进行预测,将各分量预测结果叠加后得到最终风速预测值。通过C++语言编程进行算法实现,采用实际风场数据进行仿真,其结果表明,所提方法计算时间较短,预测精度较高,特别适用于在线超短期（10 min）和短期（1 h）的风速预测,具有实际的工程应用价值。     

基于完整自适应噪声集成经验模态分解的LSTM-Attention网络短期电力负荷预测方法

短期电力负荷预测在电网安全运行和制定合理调度计划方面发挥着重要作用。为了提高电力负荷时间序列预测的准确度,提出了一种由完整自适应噪声集成经验模态分解(complete ensemble empirical mode decomposition with adaptive noise, CEEMDAN)和基于注意力机制的长短期记忆神经网络(long short-term memory network based on attention mechanism, LSTM-Attention)相结合的短期电力负荷预测模型。完整自适应噪声集成经验模态分解有效地将负荷时间序列分解成多个层次规律平稳的本征模态分量,并通过神经网络模型预测极大值,结合镜像延拓方法抑制边界效应,提高分解精度,同时基于注意力机制的长短期记忆神经网络自适应地提取电力负荷数据输入特征并分配权重进行预测,最后各预测模态分量叠加重构后获得最终预测结果。通过不同实际电力负荷季节数据分别进行实验,并与其他电力负荷预测模型结果分析进行比较,验证了该预测方法在电力负荷预测精度方面具有更好的性能。     

Neural‐hybrid control of systems with sandwiched dead‐zones

The control of systems that have sandwiched nonsmooth nonlinearities, such as a dead‐zone sandwiched between two dynamic blocks, is addressed. An adaptive inverse control scheme using a hybrid controller structure and a neural network based inverse compensator, is proposed for such systems with unknown sandwiched dead‐zone. This neural‐hybrid controller consists of an inner loop discrete‐time feedback structure incorporated with an adaptive inverse using a neural network for the unknown dead‐zone, and an outer‐loop continuous‐time feedback control law for achieving desired output tracking. The dead‐zone compensator consists of two neural networks, one used as an estimator of the sandwiched dead‐zone function and the other for the compensation itself. The compensator neural network has neurons that can approximate jump functions such as a dead‐zone inverse. The weights of the two neural networks are tuned using a modified gradient algorithm. Simulation results are given to illustrate the performance of the proposed neural‐hybrid controller. Copyright © 2002 John Wiley & Sons, Ltd.     

基于Shapley组合模型及神经网络的电能表需求预测研究

针对电能表需求预测问题,建立基于Shapley组合模型及神经网络的电能表合理优化分配模型,以提升需求预测精度.文章通过挖掘历史数据,采用Holt-Winters、BP神经网络和RBF神经网络模型对电能表需求分别进行预测、对比和分析,并且引入Shapley法对三类预测模型进行组合建模,求取相应模型的权重,获取最优的生产调度方案.仿真实验结果表明,RBF神经网络模型预测精度要高于BP神经网络和Holt-Winters模型.相较于单一模型,Shapley法组合模型具有更好的效果和实用性,有助于国家电网公司建立高效、科学的生产调度计划.     

适用于小样本的神经网络光伏预测方法

基于神经网络的短期光伏预测方法通常需要大量训练样本,对于新投运的光伏电站,历史运行数据的不足使得常规短期光伏预测方法难以应用。针对该问题,提出一种适用于小样本的双层神经网络单步光伏预测方法。根据光伏发电各环节影响因素的解耦特性,将常规单层神经网络拆分为双层网络,使每层网络具有简化的结构;用单步预测代替多步预测,降低神经网络的输入输出维数;基于统计分析,将天气影响因素有效整合到预测模型中,简化输入输出之间的映射关系。使用实际数据对所提光伏预测模型进行训练和验证,结果表明,所提方法可有效减少对训练样本数量的需求,同时保证预测的准确度。     

基于CEEMD-GRU模型的短期电力负荷预测方法

针对电力负荷序列不平稳、随机性强,直接输入模型会导致拟合效果差、预测精度低等问题,本文提出了一种基于添加互补白噪声的互补集合经验模态分解(complementary ensemble empirical mode decomposition, CEEMD)以及门控循环单元神经网络(gated recurrent unit neural network, GRU)融合的预测方法。首先,针对传统经验模态分解(empirical mode decomposition, EMD)分解方法处理干扰信号大的序列时,存在的模态混叠问题,提出了CEEMD分解方法,加入互补白噪声,将原始序列分解成不同尺度的子序列,随后使用GRU神经网络,并优化网络超参数,从而获得最好的预测结果。通过实验证明,该方法重构误差小,预测效果好。     

基于遗传算法优化BP神经网络的接触电阻预测

接触电阻是反应导体间电接触性能的重要参数,在实际的工程中往往采用经验公式对接触电阻进行计算,精度难以满足要求。为解决这一问题,将遗传算法(GA)与BP神经网络相结合对接触电阻进行预测。通过实验得到数据,分别利用遗传算法优化BP神经网络、BP神经网络以及回归分析模型进行训练和测试,并将各算法所得误差进行对比。误差对比结果表明：遗传算法优化BP神经网络的收敛效果优于其他两种算法,且遗传算法优化BP神经网络所得接触电阻模型的相对误差平均值比BP神经网络减少了4.01%,比回归分析减少了4.72%,且预测效果较稳定。利用遗传算法与BP神经网络相结合的接触电阻预测模型较单独使用BP神经网络预测模型具有更好的非线性拟合能力和更高的预测精度。     

Precision motion control of a small launching platform with disturbance compensation using neural networks

A kind of launching platform driven by two permanent magnet synchronous motors which is used to launch kinetic load to hit the target always faces strong parameter uncertainties and strong external disturbance such as the air current impulsion which would degrade their tracking accuracy greatly. In this paper, a practical method which combines adaptive robust control with neural network‐based disturbance observer is proposed for high‐accuracy motion control of the launching platform. The proposed controller not only accounts for the parametric uncertainties but also takes the external disturbances into account. Adaptive control is designed to compensate the former, while neural network‐based disturbance observer is designed to compensate the latter respectively and both of them are integrated together via a feedforward cancellation technique. A new kind of parametric adaptation and weight adaptation strategy is designed by using the linear combination of the system's tracking error and the weight estimation error as a driving signal for parametric adaptation and disturbance approximation. The stability of the novel control scheme is analyzed via a Lyapunov method and this method presents a prescribed output tracking performance in the presence of both parameter uncertainties and unmodeled nonlinearities. Extensive comparative simulation and experimental results are obtained to verify the high‐performance of the proposed control strategy. Copyright © 2016 John Wiley & Sons, Ltd.     

微风速流场数据采集与融合方法研究

在微风速(0～1 m/s)空间流场测量中,对传感器精度要求高,实时在线仪表数据精度不够,数据采集滞后性大;考虑采用多个传感器测量提高精度,但也存在数据融合的问题.针对微风速流场测量,提出基于K均值RBF神经网络的数据采集预处理软测量模型,首先选取中间变量(电流值),运用K均值聚类,用RBF网络训练得到单个传感器数据;提出基于相关性kalman滤波的传感器数据融合算法,剔除无效数据点,并融合得到精确风速预测值.测量实验和数据结果表明该方法处理的数据结果滞后性小,处理速度快,数据精度高.     

数据融合在物联网火灾监测系统中的应用研究

利用数据融合技术,可以弥补单一传感器数据所造成的结果不完整以及片面的缺陷。为了保证火灾判断的准确性,提出神经网络和模糊理论相结合的火灾监测算法,并将算法应用在火灾监测系统中。通过利用神经网络工具箱以及模糊推理工具箱,对两种算法分别进行了MATLAB仿真和分析,得出将两种算法结合适用于火灾监测的结论。最后在火灾监测开发平台上利用VS2010实现对火灾数据的处理,得出火灾发生概率,而后判断火灾发生的可能性,实现了火灾监测算法功能,保证了判断的准确性。     

基于NAR动态神经网络的BDS周跳探测与修复方法

针对北斗导航定位系统(BDS)数据处理过程中出现的周跳问题,提出一种提升小波结合NAR动态神经网络的周跳探测与修复方法。首先构造了非差周跳检验量,通过提升小波法探测到周跳发生历元,再采用NAR动态神经网络法、改进BP神经网络法以及传统多项式拟合法,分析对比不同方法周跳修复效果。实验仿真结果表明,在周跳探测方面,提升小波法可有效探测0.2周以上的小周跳;在周跳修复方面,NAR神经网络比改进BP神经网络的拟合度提高40%左右,预测精度比改进的BP神经网络提高50%左右,比传统多项式拟合法提高10%以上,更适用于小周跳的探测与修复,进一步提高了定位精度。     

基于数据分析和改进Chebyshev神经网络的风速时间序列预测

为提高风速时间序列预测精度,基于风速时间序列的随机性和波动性,提出互补集合经验模态分解(Complete Ensemble Empirical Mode Decomposition,CEEMD)和正交粒子群算法(Orthogonal Particle Swarm Optimization,OPSO)优化Chebyshev基函数神经网络的混合风速时间序列预测模型(CEEMD-OPSO-Chebyshev)。利用CEEMD将原始风速时间序列分解成有限个固有模态分量,避免了传统的分解信号重建中冗余噪声残留问题。同时引入排列熵分析各分量内在特性进行聚类,提出基于OPSO优化算法的Chebyshev神经网络风速预测模型,利用OPSO优化预测网络权值,进一步提高预测精度,通过对实际采样的风电场风速时间序列进行预测分析,结果可得所提出的混合预测模型与传统预测模型相比能得到更高的预测精度。     

关于 PDR 算法步长估计模型的改进研究

为了更加精确地判别基于微惯性测量单元( IMU)的行人定位信息,本文深入研究了传统行人航迹推算(PDR)算法模 型,发现传统算法所采用的判别条件单一且精准度不高。 针对传统算法中步长估计模型不准确的问题,本研究首先提出一种基 于扩展卡尔曼滤波的误差补偿优化算法,以实现 IMU 内集成的加速度计、陀螺仪等传感器的误差补偿。 将优化后的原始数据 放入 BP 神经网络算法对单参数步长估算经验模型进行训练。 实验结果表明,基于 BP 神经网络融合基础模型的步长算法相比 单纯的基础步长模型,闭环精度提高了 0. 3%以上,开环误差减小了 8. 5 倍,基于 BP 神经网络的改进 PDR 算法可以有效抑制惯 性算法的误差发散。     

A novel probabilistic neural network‐based algorithm for classifying internal fault in transformer windings

The major function of protective devices in a power system is to detect the occurrence of faults and to isolate the faulty sections from the rest of the system. Much progress has been made in the development algorithms for detecting faults in power transformers, which depend on transients‐based techniques. This paper presents an algorithm based on a combination of discrete wavelet transforms and probabilistic neural networks (PNNs) for classifying internal faults in a two‐winding three‐phase transformer. Fault conditions of the transformer are simulated using alternative transients program/electromagnetic transients program (ATP/EMTP) in order to obtain current signals. The mother wavelet Daubechies4 is employed to decompose the high‐frequency components from these signals. All three phases of the differential current signals are used in the fault detection decision algorithm. The variations of first‐scale high‐frequency component that detects fault are used as an input for the training pattern. The training process for the neural network and fault diagnosis decision is implemented using toolboxes on MATLAB/Simulink. Various cases and fault types based on the Thailand electricity transmission and distribution systems are studied to verify the validity of the algorithm. Backpropagation neural network is also compared with the PNN in this paper. It is found that the proposed method gives satisfactory accuracy with less training time, and will be particularly useful in the development of a modern differential relay for a transformer protection scheme. © 2013 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.     

基于GA优化BP神经网络的液压钻机故障诊断

针对BP神经网络应用于故障诊断时存在着收敛速度慢、易陷入局部最小值等问题,提出了一种基于遗传算法(GA)优化BP神经网络的液压钻机故障诊断方法.利用GA的选择、交叉和变异操作优化BP神经网络的权值和阈值,提高网络训练的收敛速度.根据液压钻机工况参数提取的特征信号,进行归一化处理建立样本,利用训练样本对网络进行训练,根据训练结果进行故障诊断.仿真结果表明,GA优化的BP神经网络迭代次数少,收敛速度快,能够对测试样本进行有效地分类,故障诊断正确率高.     

Automatic EEG interpretation adaptable to individual electroencephalographer using artificial neural network

In view of the fact that the visual EEG interpretation could be a subjective task and may vary among the electroencephalographers, the main objective of this study was to develop an automatic EEG interpretation system which is adaptable to each electroencephalographer. The system adapted to each electroencephalographer would bring a close automatic EEG interpretation to that done by the electroencephalographer's visual interpretation. The adaptable automatic EEG interpretation was accomplished by using the constructive neural network with forgetting factor. The artificial neural network was constructed so as to give the integrative interpretation of the EEG based on the intermediate judgment of 13 items that characterized the visual interpretation. The developed method was evaluated based on the visually inspected EEG data of 37 patients by electroencephalographer‐A and the data of 20 patients by electroencephalographer‐B. The adapted ANN showed good agreement with each electroencephalographer's visual inspection. The proposed automatic EEG interpretation by use of the ANN can be a powerful assistant tool for individual electroencephalographers for their EEG interpretation. Copyright © 2001 John Wiley & Sons, Ltd.