Computational Intelligence Prediction Model Integrating Empirical Mode Decomposition,Principal Component Analysis,and Weighted k-Nearest Neighbor

On the basis of machine leaning,suitable algorithms can make advanced time series analysis.This paper proposes a complex k-nearest neighbor(KNN)model for predicting financial time series.This model uses a complex feature extraction process integrating a forward rolling empirical mode decomposition(EMD)for financial time series signal analysis and principal component analysis(PCA)for the dimension reduction.The information-rich features are extracted then input to a weighted KNN classifier where the features are weighted with PCA loading.Finally,prediction is generated via regression on the selected nearest neighbors.The structure of the model as a whole is original.The test results on real historical data sets confirm the effectiveness of the models for predicting the Chinese stock index,an individual stock,and the EUR/USD exchange rate.

Parallel Active Subspace Decomposition for Tensor Robust Principal Component Analysis

Tensor robust principal component analysis has received a substantial amount of attention in various fields.Most existing methods,normally relying on tensor nuclear norm minimization,need to pay an expensive computational cost due to multiple singular value decompositions at each iteration.To overcome the drawback,we propose a scalable and efficient method,named parallel active subspace decomposition,which divides the unfolding along each mode of the tensor into a columnwise orthonormal matrix(active subspace)and another small-size matrix in parallel.Such a transformation leads to a nonconvex optimization problem in which the scale of nuclear norm minimization is generally much smaller than that in the original problem.We solve the optimization problem by an alternating direction method of multipliers and show that the iterates can be convergent within the given stopping criterion and the convergent solution is close to the global optimum solution within the prescribed bound.Experimental results are given to demonstrate that the performance of the proposed model is better than the state-of-the-art methods.

Assessment of depth of anesthesia using principal component analysis

A new approach to estimating level of uncon-sciousness based on Principal Component Analysis (PCA) is proposed. The Electroen-cephalogram (EEG) data was captured in both Intensive Care Unit (ICU) and operating room, using different anesthetic drugs. Assuming the central nervous system as a 20-tuple source, window length of 20 seconds is applied to EEG. The mentioned window is considered as 20 nonoverlapping mixed-signals (epoch). PCA algorithm is applied to these epochs, and larg-est remaining eigenvalue (LRE) and smallest remaining eigenvalue (SRE) were extracted. Correlation between extracted parameters (LRE and SRE) and depth of anesthesia (DOA) was measured using Prediction probability (PK). The results show the superiority of SRE than LRE in predicting DOA in the case of ICU and isoflurane, and the slight superiority of LRE than SRE in propofol induction. Finally, a mixture model containing both LRE and SRE could predict DOA as well as Relative Beta Ratio (RBR), which expresses the high capability of the proposed PCA based method in estimating DOA.

Beam position monitor troubleshooting by using principal component analysis in Shanghai Synchrotron Radiation Facility

Beam position monitors(BPMs)have been widely used in all kinds of measurement systems,feedback systems and other areas in particle accelerator field these days.The malfunction of a single BPM can cause serious consequences such as the failure of the orbit feedback and the transverse feedback.A troubleshooting has been made to prevent the defective BPMs from affecting the accuracy and stability of the storage ring in Shanghai Synchrotron Radiation Facility(SSRF).Different types of malfunctions have been successfully identified by using the idea of principal component analysis(PCA).

Identification of the anomaly component using BEMD combined with PCA from element concentrations in the Tengchong tin belt, SW China

Concentration of elements or element groups in a geological body is the result of multiple stages of rockforming and ore-forming geological processes.An ore-forming element group can be identified by PCA(principal component analysis)and be separated into two components using BEMD(bi-dimensional empirical mode decomposition):(1)a high background component which represents the ore-forming background developed in rocks through various geological processes favorable for mineralization(i.e.magmatism,sedimentation and/or metamorphism);(2)the anomaly component which reflects the oreforming anomaly that is overprinted on the high background component developed during mineralization.Anomaly components are used to identify ore-finding targets more effectively than ore-forming element groups.Three steps of data analytical procedures are described in this paper;firstly,the application of PCA to establish the ore-forming element group;secondly,using BEMD on the o re-forming element group to identify the anomaly components created by different types of mineralization processes;and finally,identifying ore-finding targets based on the anomaly components.This method is applied to the Tengchong tin-polymetallic belt to delineate ore-finding targets,where four targets for Sn(W)and three targets for Pb-Zn-Ag-Fe polymetallic mineralization are identified and defined as new areas for further prospecting.It is shown that BEMD combined with PCA can be applied not only in extracting the anomaly component for delineating the ore-finding target,but also in extracting the residual component for identifying its high background zone favorable for mineralization from its oreforming element group.

Structural Damage Detection Method Based on Decomposition of the Operating Deflection Shapes

Full-field measurement techniques such as the scanning laser Doppler vibrometer （LDV） and the electronic speckle pattern interferometry systems can provide a dense and accurate vibration measurement on structural operating deflection shape （ODS） on a relatively short period of time.The possibility of structural damage detection and localization using the ODS looks likely more attractive than when using traditional measurement techniques which address only a small number of discrete points.This paper discusses the decomposition method of the structural ODSs in the time history using principal component analysis to provide a novel approach to the structural health monitoring and damage detection.The damage indicator is proposed through comparison of structural singular vectors of the ODS variation matrixes between the healthy and damaged stages.A plate piece with a fix-free configuration is used as an example to demonstrate the effectiveness of the damage detection and localization using the proposed method.The simulation results show that：（1） the dominated principal components and the corresponding singular vectors obtained from the decomposition of the structural ODSs maintain most of all vibration information of the plate,especially in the case of harmonic force excitations that the 1st principal component and its vectors mostly dominated in the system;（2） the damage indicator can apparently flag out the damage localization in the case of the different sinusoidal excitation frequencies that may not be close to any of structural natural frequencies.The successful simulation indicates that the proposed method for structural damage detection is novel and robust.It also indicates the potentially practical applications in industries.

Distributed Monitoring of Power System Oscillations Using Multiblock Principal Component Analysis and Higher-order Singular Value Decomposition

The primary goal in the analysis of hierarchical distributed monitoring and control architectures is to study the spatiotemporal patterns of the interactions between areas or subsystems.In this paper,a novel conceptual framework for distributed monitoring of power system oscillations using multiblock principal component analysis(MB-PCA)and higher-order singular value decomposition(HOSVD)is proposed to understand,characterize,and visualize the global behavior of the power system.The proposed framework can be used to evaluate the influence of a given area or utility on the oscillatory behavior,uncover low-dimensional structures from high-dimensional data,and analyze the effects of heterogeneous data on the modal characteristics and interpretation of power system.The metrics are then investigated to examine the relationships between the dynamic patterns and participation of individual data blocks in the global behavior of the system.Practical application of these techniques is demonstrated by case studies of two systems:a 14-machine test system and a 5449-bus 635-generator equivalent model of a large power system.

Comparative Analysis of Velocity Decomposition Methods for Internal Combustion Engines

Different signal processing technique performances are compared to each other with regard to separating the mean and fluctuating velocity components of a simulated one-dimensional unsteady velocity signal comparable to signals observed in internal combustion engines. A simulation signal with known mean and fluctuating components was generated using experimental data and generic turbulence spectral information. The simulation signal was generated based on observations on the measured velocity data obtained using LDV in a motored Briggs-and-Stratton engine at about 600 RPM. Experimental data was used as a guide to shape the simulated signal mean velocity variation;fluctuating velocity variations with specified spectrum and standard deviation was used to mimic the turbulence. Cyclic variations were added both to the mean and the fluctuating velocity signals to simulate prescribed cyclic variations. The simulated signal was introduced as input to the following algorithms: ensemble averaging;high-pass filtering;Proper-Orthogonal Decomposition (POD);Wavelet Decomposition (WD) and Wavelet Decomposition/Principal Component Analysis (WD/PCA). The results were analyzed to determine the best method in correctly separating the mean and the fluctuating velocity information, indicating that the WD/PCA performs better compared to other techniques.

白条猪价格预测模型构建

【目的】增强农产品价格预测准确度,为农产品价格的有效预测提供参考。【方法】以河南省白条猪每周平均批发价格为研究对象,提出一种基于序列分解、主成分分析和神经网络(CEEMDAN-PCA-CNN-LSTM)的白条猪价格预测方法。首先,使用自适应白噪声完全集合模态分解方法(CEEMDAN)对白条猪价格序列进行分解;其次,选用皮尔逊相关系数筛选影响价格波动的相关因素;再次,利用主成分分析(PCA)对影响因素及分解得到的子序列降维处理并作为原始价格序列的特征值,并行输入到作为编码器的卷积神经网络(CNN)中进行特征提取;最后,引入长短期记忆网络(LSTM)作为解码器输出得到预测结果。将该方法应用于河南省白条猪每周平均价格数据,与LSTM、门控循环单元(GRU)、CNN、基于卷积的长短期记忆网络(ConvLSTM)模型进行比较。【结果】CEEMDAN-PCA-CNN-LSTM组合模型预测方法得到的平均绝对误差分别降低了44.95%、27.30%、28.13%、43.17%。【结论】CEEMDAN-PCA-CNN-LSTM模型对于河南省白条猪市场价格的预测性能更优,有助于相关部门针对河南省白条猪价格波动做出科学决策。

基于层次分解、主成分分析和高斯混合模型的火成岩岩性识别——以惠州26洼古潜山为例

火成岩油气成藏规律复杂,受到火山运动、强构造运动以及风化剥蚀等叠加影响,火成岩的化学成分和结构构造复杂多样,非均质性极强,采用常规岩性识别方法难以一次性将所有岩性准确识别。借鉴层次分解思路,以惠州26洼古潜山为例,提出了一种火成岩岩性测井识别分类方法:综合考虑火成岩地质分类原则和测井响应特征来确定岩性识别层级,基于这种层次性的分类原则,在每一层次定量优选岩性识别敏感参数,建立研究区岩性识别优选层级;在明确岩性识别优选层级的基础上,逐级逐次使用主成分分析(PCA)和高斯混合模型(GMM)对岩性进行判别并确定其计算函数,建立分级优选岩性识别模型,最终达到整体岩性区分的目的。研究结果表明,研究区辉绿岩和闪长岩识别正确率分别为87.31%和84.32%,未分级未优选辉绿岩和闪长岩识别正确率为60.45%和54.88%,分级未优选其岩性识别正确率为69.61%和67.04%,有效提高了研究区的复杂岩性识别精度。该方法的提出对提高火成岩岩性识别精度提供了一种思路,也为研究区古潜山火成岩岩性精确识别提供了参考依据。

基于经验模态分解和深度学习的短期风电功率预测

精准的风电功率预测有利于全网电力平衡、系统安全稳定运行和节能减耗。提出一种基于经验模态分解(empirical mode decomposition, EMD)、核主成分分析(kernel principal component analysis, KPCA)和长短期记忆(long short-term memory, LSTM)神经网络的短期风功率预测模型。采用EMD技术将多维气象序列分解为多个固有模态分量,以挖掘原始数据的主要特征并消除噪声;引入KPCA进行降维处理,提取数据的非线性特征;使用LSTM神经网络对特征提取的序列进行学习并完成预测,获得风电功率预测的最终结果。使用所提出的模型对新疆某一风电场风电功率进行预测,将预测结果与其他模型对比。结果表明,该预测模型能改善预测性能,降低风电功率预测误差。

中国出口集装箱运价指数影响因素的实证研究

2020年,新型冠状病毒感染的流行,中国出口集装箱运价指数(CCFI)“疯涨”且不断创造历史新高。首先运用集合经验模态分解法将CCFI序列分解为若干分量,重构后得到高频部分、低频部分和趋势项,其次采用主成分分析法从选取的8个宏观影响因素中提取出两个公共因子,定义为国际贸易环境和中国出口形势,分别研究不同频率下的CCFI指数与影响因素之间的关系,探究病毒流行前后CCFI指数影响因素的变化。结果显示:CCFI的短期波动项在感染前不受宏观因素影响,感染流行后表现为正相关;病毒流行后CCFI的重大事件影响项与国际贸易环境由负相关变成正相关,与中国出口形势始终保持正相关;CCFI的长期趋势项与宏观因素在病毒流行前后均呈现正相关,且在病毒流行后正相关程度有所提高。

基于EMD-PCA-LSTM的短期风电功率预测研究

精准的风电功率预测有利于全网电力平衡、系统安全稳定运行和节能减耗,提出一种基于EMD-PCA-LSTM的短期风电功率预测模型。先采用经验模态分解技术将多维气象序列分解为多个固有模态分量,以挖掘原始数据的主要特征并消除噪声。再引入主成分分析进行降维处理,提取数据的非线性特征,最后使用长短期记忆神经网络进行预测。通过与多种预测模型进行比较,证明了该模型在预测精度方面的卓越表现。

基于EEMD和特征降维的非侵入式负荷分解方法研究

针对现有非侵入式居民用电负荷监测缺乏对独立负荷完整、全面的分解方法,导致用电信息的完整性得不到保证的不足,提出一种基于集合经验模态分解(ensemble empirical mode decomposition,EEMD)和Pearson-PCA改进的盲源分离算法。利用EEMD对总功率信号分解,以消除经验模态在分解过程中易出现模态混叠的现象,并得到一系列固有模式函数(intrinsic mode functions,IMF)。结合Pearson相关系数和主成分分析法(principal component analysis,PCA),提出Pearson-PCA改进算法对IMF进行降维,剔除相关性较弱的IMF分量,以及估计源信号数目。运用快速独立分量分析(fast independent component analysis,FastICA)对降维后的IMF进行分解,计算得出源功率信号。将提出的改进算法应用于非侵入式居民用电负荷分解问题,采用能量分解数据集(reference energy disaggregation data,REDD)进行实验仿真。实验结果表明:在不同用电场景下,提出的改进算法均具有较好的分解效果。

基于同态加密的隐私保护主成分分析方法

在现实生活中,不同的行业之间,甚至同行业不同部门之间的数据并不互通,随着计算机算力的提升,制约模型训练效果的不是算力而是数据量。因此,想要得到更好的算法模型,仅靠某一方的数据是不够的,需要两方或者多方的参与,这就要求对各方的数据进行隐私保护。除此之外,随着收集的数据越来越详细,数据的维数也越来越大。面对高维的数据,数据降维是不可缺少的环节,而在数据降维方面,主成分分析(Principal Component Analysis,PCA)是常用的手段。当拥有数据的两方想要合作进行隐私保护的数据降维时,同态加密技术是一种解决办法。同态加密技术可以在保护数据隐私的前提下对加密数据进行计算,可以用在加密数据的PCA上。针对上述应用场景,利用CKKS同态加密方案,通过幂法迭代的SVD技术设计了一种两方加密数据进行PCA的方案,在保护两方数据隐私的前提下实现数据降维的目的;通过改进传统幂法迭代步骤,避免了代价高昂的同态密文除法运算,使得在选取较小的加密参数时,也能支持更多的幂法迭代次数,从而在缩短同态计算时间的同时提高计算精度。在公共数据集上进行测试,并与现有方案进行对比,该方案在计算耗时上缩短了约80%,与明文计算结果的均方误差缩减到1%以内。

基于多源数据融合的分布式光伏聚合超短期预测方法

分布式光伏聚合发电的超短期预测是支撑其功率快速调节的前提保障,由于规模化接入的分布式光伏容量小、分布广,其发电时序特性差异性大、非平稳性强,导致其超短期预测精度难以保证。为此,文章提出基于多源数据融合的分布式光伏聚合超短期预测方法。该方法基于变分模态分解法,充分挖掘分布式光伏聚合发电非平稳性特性,并采用核主成分分析法对引发光伏发电非平稳性的影响因素即温度、湿度、光照、云量等多源数据进行量化解析,同时结合改进的长短期记忆神经网络,创建了多源数据融合方法,实现了分布式光伏聚合发电超短期预测。仿真结果表明,该方法有效提升了模型的预测精度。与传统方法相比,提出的预测方法对随机性波动严重的光伏超短期预测具有显著优势。

变频器负载回路串联故障电弧检测及选线方法

串联故障电弧的高温是引发电气火灾的主要原因之一,针对工业变频器负载回路中串联故障电弧尚无有效保护手段的问题,提出了一种新的串联故障电弧检测及选线方法。首先,针对工业领域常用的三相变频器负载回路开展了不同线路中发生串联故障电弧的实验;其次,利用基于能量收敛原则改进的变分模态分解将变频器前端A相电流信号自适应分解为多个模态分量,依次将单个模态分量乘以能量系数并重构,得到多个电流信号的特征增强信号,并建立特征矩阵;再次,对特征矩阵进行分块,利用核主成分分析对每块矩阵进行降维,并对降维信号组成的矩阵进行二次降维构建故障特征向量;最后,利用鹈鹕算法优化的支持向量机对串联故障电弧进行检测及选线。结果表明:该方法仅通过分析变频器前端A相电流可以实现变频器整个回路中6条线路的串联故障电弧检测及选线,检测及选线准确率均达到98%以上。

基于PCA和EEMD的柔性直流配电网故障选线算法

柔性直流故障选线技术的发展对直流配电网有着至关重要的作用。本文针对现有柔性直流故障存在的可利用的故障信息较少等问题,提出了一种新算法,该算法有效利用了集合经验模态分解EEMD(ensemble empirical mode decomposition)算法、主成分分析PCA(principal component analysis)和相关系数各自的优势。首先,提取暂态电流样本信号,采用EEMD得到以正交基函数表示的数据矩阵;接着,基于PCA进行该矩阵元素特征向量到主成分的转换,将样本信号投影到主元空间实现坐标变换,从而得到对样本数据的聚类和识别结果;最后,基于相关系数进行故障线路判别。本文算法的EEMD揭露了原始历史数据的内在变化规律,PCA能够有效选择故障有效特征。大量实验表明,该新算法准确有效,与现有其他方法相比,在故障信息不明显、不同过渡电阻方面具有优势。

一种融合KPCA、FastICA及SVD的腹壁源胎儿心电 信号提取算法研究

目的:为实现从母体腹壁混合信号中提取高信噪比和波形清晰的胎儿心电信号,提出一种融合核主成分分析(kernel principal component analysis,KPCA)、快速独立成分分析(fast independent component analysis,FastICA)及奇异值分解(singular value decomposition,SVD)的胎儿心电信号提取算法。方法:首先,采用KPCA对母体心电信号进行降维,再利用改进的基于负熵的FastICA处理降维后的数据,得到独立成分。随后,引入样本熵进行信号通道选择,挑选出包含最多母体信息的信号通道。在选中的母体通道上进行SVD,得到母体心电信号的近似估计,再用腹壁源信号减去该信号得到胎儿心电的初步估计。最后,采用改进的基于负熵的FastICA成功分离出纯净的胎儿心电信号。在腹部和直接胎儿心电图数据库(Abdominal and Direct Fetal Electrocardiogram Database,ADFECGDB)和PhysioNet 2013挑战赛数据库中对提出的算法进行验证。结果:提出的算法在主观视觉效果和客观评价指标上都表现出优越的性能。在ADFECGDB数据库中,胎儿QRS复合波检测的敏感度、阳性预测值和F1值分别为99.74%、98.85%和99.30%;在PhysioNet 2013挑战赛数据库中,胎儿QRS复合波检测的敏感度、阳性预测值和F1值分别为99.10%、97.87%和98.48%。结论:融合KPCA、FastICA及SVD的胎儿心电信号提取算法在提取胎儿心电信号的同时有效处理了附加噪声,为胎儿疾病的早期诊断提供了有力支持。

基于数据驱动的离心泵轴承特征分析及寿命预测

离心泵是工业中能量转换和流体输送的核心设备,其部件滚动轴承的可靠性对整个机组的安全运行尤为关键。为了解决目前滚动轴承寿命预测问题,对滚动轴承剩余寿命的最佳预测方案进行了研究。首先,从数据驱动和试验出发,利用试验台采集所得的离心泵轴承正常及故障状态下的数据,分析了时域、频域、时频域各特征在不同工况中的表现差异,发现了时域特征、频域特征、小波包分解能量特征、完全自适应噪声完备集合经验模态分解(CEEMDAN)能量特征可以捕捉到不同工况下的故障信息;然后,以单调性、趋势性指标加权分数为依据,结合特征的敏感性分析结果,优选出了轴承在全寿命周期中表现突出的12个特征,经核主成分分析(KPCA)-长短期记忆网络(LSTM)降维处理后,构建出了能够表征离心泵轴承退化过程的一维特征量;最后,对比分析了LSTM网络、反向传播(BP)网络和卷积神经(CNN)网络的预测效果。研究结果表明:LSTM网络的均方根误差(RMSE)为0.402,平均绝对百分比误差(MAPE)为0.332,预测精度在三者中最好,模型平均训练时间为12.6 s,可见LSTM网络在预测精度及模型训练时间上更具优势。