基于SVM算法的电力物资采购价格预测研究

电力工程建设涉及物资种类繁多,物资价格受很多因素的影响,尽管当前许多学者已经提出关于电力物资采购价格的预测方法,但是多数模型局限于单因素考虑,鲜有研究对电力物资采购价格进行多因素综合分析。为更合理地预测当下电力物资价格走势并建立电力价格预测模型,该文首先针对收集的历史采购数据,总结影响电力物资采购价格的核心因素,其次通过支持向量机（SVM）算法对物资价格进行多因素分析预测,并与神经网络算法进行比较,结果表明,采用支持向量机的算法模型的预测结果与实际情况吻合度较好且比神经网络算法误差更小,可为电力公司物资采购提供参考。     

基于SVM的短期风电功率预测

该文针对风电功率的强非线性、大波动性特点,提出一种支持向量机（SVM）短期风电功率预测方法,并给出了具体的建模过程和流程。最后选用国内某风电场的实际运行数据进行验证,并与传统的BP神经网络算法进行了对比,结果表明：该文提出的风电功率预测方法能更好地跟踪风电功率的变化,而且该预测模型的预测精度更高,能更好地提供风电功率预测数据。     

ESVR算法及支撑向量机算法中参数的确定

支撑向量机是基于有限数据的机器学习算法,主要研究如何从一些给定的观测数据获得目前尚不能通过原理分析得出的规律,利用这些规律去分析客观现象并对无法观测的数据进行预测。本文在已有的支撑向量机算法的基础上,提出了一种新的算法——ESVR算法,它是基于支撑向量回归机的改进算法,利用原有用于回归问题的SVM算法消除了孤立点对已知问题的影响。针对支撑向量机算法中核参数取值对推广性的影响较明显的特点,本文给出了一种核函数中参数的确定方法——渐进搜索法,它可以得到支撑向量机算法中核参数的取值范围,并具有推广误差较小的特点。数值实验表明它们具有较好的效果。     

基于差分进化支持向量机的轧制力预测

板带材热连轧轧制过程中,轧制力的精确控制对改善带钢板形性能有至关重要的作用.针对B钢厂2250热连轧轧线轧制力模型计算值与实际值误差较大(±10％)的问题,建立基于线性(Linear)核函数、多项式(Poly)核函数、高斯(RBF)核函数3种支持向量机回归预测模型,分析3种核函数支持向量机预测模型,选用拟合效果最好的R...     

最小二乘支持向量机的扩展及其在时间序列预测中的应用

根据时间序列近期数据较远期数据包含有更多未来信息的思想,对最小二乘支持向量机预测方法进行了扩展,得到了更具一般性的最小二乘支持向量机预测模型,给出了扩展后的预测模型具体算法。两个时间序列的预测实例表明,扩展后的预测方法获得了更好的预测效果,提升了最小二乘支持向量机预测方法的价值。     

基于混沌时间序列的GPS单点定位误差预测方法

对由不同型号GPS-OEM板和不同型号天线所组成的GPS单点定位系统的定位误差时间序列进行了混沌特性分析,通过计算最大Lyapunov指数确定了其为混沌时间序列.在此基础上提出了基于混沌时间序列的GPS单点定位误差预测方法,分别将RBF神经网络和Volterra滤波器应用于预测模型,实现了对定位误差的一步预测.通过对比预测结果,发现Volterra滤波器比RBF神经网络具有更高的预测精度,且算法实时性强,可以有效地降低预测模型的复杂程度.     

层次型支持向量机人脸检测器

大量的数值实验表明，在非线性支持向量机中，核函数的选取对支持向量机性能的影响很大。核函数的选择一直是一个难题。本文利用遗传算法的全局优化能力比较并分析了多项式核函数和高斯核函数的检测正确率和支持向量个数两项指标，提出了一种层次型支持向量机人脸检测器。实验结果证明，本文的方法确实取得了较好的效果。     

组合核函数支持向量机在水中目标识别中的应用

论文研究了支持向量机核函数构成条件以及不同核函数的特性，结合水中目标识别技术特点，提出了一种组合核函数支持向量机的方法。提取了基于小波变换的舰船辐射噪声奇异性、尺度-过零、尺度-能量特征，对水中目标进行了SVM分类识别。研究表明，基于组合核函数的支持向量机分类识别效果优于单独核函数的支持向量机识别效果。     

SOM—SVM模型在煤与瓦斯突出预测中的应用

将自组织特征映射网络和支持向量机进行优选组合,建立煤与瓦斯突出危险性预测的SOM—SVM模型,充分利用非监督学习算法SOM的数据压缩、特征抽取的功能特性对训练样本进行压缩去噪处理,为有导师学习算法SVM提供高质量的有标记样本,进而发挥SVM分类精度高的特性,同时提高其分类速率。通过现场实测数据进行煤与瓦斯突出危险性预测,结果表明：两种算法的结合对煤与瓦斯突出危险性预测是有效的,它与传统的预测方法相比,分类速度更快,容错能力更强,预测精度更高。     

新能源功率预测算法优化研究

以风能和太阳能为代表的新能源具有随机性、间歇性和波动性,对新能源发电功率进行预测是有效解决以上问题的途径。在确定性预测中充分考虑风电出力和预测模型特性,提出分段支持向量机（piecewise support vector machine,PSVM）和神经网络（neural network,NN）预测算法;充分考虑天气特征对光伏出力的影响,提出基于气象特性分析的光伏出力预测算法。通过若干风电场的算例分析,证明了上述几种预测模型的实用性,为功率预测的可靠性分析提供支持。     

Failure and reliability prediction by support vector machines regression of time series data

Support Vector Machines (SVMs) are kernel-based learning methods, which have been successfully adopted for regression problems. However, their use in reliability applications has not been widely explored. In this paper, a comparative analysis is presented in order to evaluate the SVM effectiveness in forecasting time-to-failure and reliability of engineered components based on time series data. The performance on literature case studies of SVM regression is measured against other advanced learning methods such as the Radial Basis Function, the traditional MultiLayer Perceptron model, Box-Jenkins autoregressive-integrated-moving average and the Infinite Impulse Response Locally Recurrent Neural Networks. The comparison shows that in the analyzed cases, SVM outperforms or is comparable to other techniques.     

基于DT-CWT和SVM的纹理分类算法

提出了一种基于双树复数小波变换(DT-CWT)和支持向量机(SVM)的纹理分类算法.双树复数小波变换不仅具有实数小波的诸多优点,而且还具有近似平移不变性、良好的方向选择性和低冗余度,并且能对图像进行完全重构,能够更好地刻画纹理的特性;支持向量机算法是近年发展起来的性能优越的分类算法,比传统分类器有很大的优越性:避免了局部最优解和"维数灾"问题,其最优分类超平面的思想能够提高分类准确度.该方法用双树复数小波对纹理图像进行滤波并在各方向子带上进行重构,再计算其局部能量函数得到每个像素的特征向量,最后利用支持向量机算法实现对纹理图像像素的分类.将本方法与其它的分类算法进行比较,实验结果表明,提出的算法能有效地提高正确分类率.     

Phase-space prediction of chaotic time series

We report on improved phase-space prediction of chaotic time series. We propose a new neighbour-searching strategy which corrects phase-space distortion arising from noise, finite sampling time and limited data length. We further establish a robust fitting algorithm which combines phase-space transformation, weighted regression and singular value decomposition least squares to construct a local linear prediction function. The scaling laws of prediction error in the presence of noise with various parameters are discussed. The method provides a practical iterated prediction approach with relatively high prediction performance. The prediction algorithm is tested on maps (Logistic, Hénon and Ikeda), finite flows (Rössler and Lorenz) and a laser experimental time series, and is shown to give a prediction time up to or longer than five times the Lyapunov time. The improved algorithm also gives a reliable prediction when using only a short training set and in the presence of small noise.     

Nonparametric factor analysis of residual time series

We introduce a nonparametric smoothing procedure for nonparametric factor analysis of multivariate time series. Our main objective is to develop an adaptive method for estimating a time-varying covariance matrix. The asymptotic properties of the proposed procedures are derived. We present an application based on the residuals from the Fair macromodel of the U.S. economy. We find substantial evidence of time varying second moments and breaks in the contemporaneous correlation structure during the mid 1970's to the early 1980's.     

Count time series prediction using particle filters

Non‐Gaussian dynamic models are proposed to analyse time series of counts. Three models are proposed for responses generated by a Poisson, a negative binomial, and a mixture of Poisson distributions. The parameters of these distributions are allowed to vary dynamically according to state space models. Particle filters or sequential Monte Carlo methods are used for inference and forecasting purposes. The performance of the proposed methodology is evaluated by two simulation studies for the Poisson and the negative binomial models. The methodology is illustrated by considering data consisting of medical contacts of schoolchildren suffering from asthma in England.     

A time integration algorithm for elasto-viscoplastic cubic crystals applied to modelling polycrystalline deformation

An algorithm for integrating the constitutive equations for an elasto-viscoplastic cubic crystal is presented which is shown to be easily employed in a polycrystalliné analysis. Anisotropic elastic behaviour is incorporated into the standard constitutive equations for ductile single crystals. The algorithm is shown to be efficient, robust and general. The primary advantage of this algorithm is that is provides an implicit integration of the plastic deformation gradient while including the elastic response. This permits taking large time steps while maintaining accuracy and stability. Several polycrystalline examples are presented to demonstrate the effect of the time step on the solution. Examples also are presented which compare the algorithm described herein to an algorithm which neglects the elastic part of the deformation. In addition, the effect of the anisotropic component of the elasticity is investigated by comparing the results with those obtained assuming isotropic elasticity.     

Corrosion of steel reinforcement due to atmospheric pollution

The aim of this study is to examine the performance of four different sets of reinforcing steel rebars, S220, S400, S500s Tempcore and S500s Vanadus, exposed to the Greek atmosphere, before their installation into the concrete. The performance against atmospheric corrosion of the aforementioned steel rebars was evaluated by means of microscopy techniques and corrosion rate measurements. In particular, we studied the influence of corrosion products of a set of reinforcement steels on the bond strength between concrete and steel bars, during the hydration process of cement. Furthermore, microscopy techniques and visual observation were used to identify the rust strains and the corrosion product morphology of steel reinforcements before their installation into the concrete. The experimental results shown that the steel type, which exhibits the higher resistance, as far as atmospheric corrosion concerns, was S220 reinforcing steel. In contrary, S500s Tempcore has the least corrosion resistance. The bond strength between concrete and the steel rebars was found to decrease with increasing weathering from 45 to 122 days, due to the morphology and to the thickness of the rust layers formed on steel surface, as observed by the ESEM.     

基于ARMA模型模拟高架桥的脉动风速时程

强风是高架桥设计与防灾减灾分析的控制性荷载之一.风与高架桥相互作用十分复杂,可以通过风洞试验、现场实测、数值模拟获取可靠的风速(风荷载)数据.尽管如此,时域分析可以使人们更全面地了解高架桥的风振响应特性, 也能更直观地反映高架桥风致振动控制的有效性.因此, 使用线性滤波法即白噪声滤波法(WNFM)中的自回归滑动平均(ARMA)模型模拟高架桥的脉动风速时程.首先, 考虑高架桥脉动风速的时间和空间相关性, 导出自回归(AR)模型阶数与滑动回归(MA)模型阶数不相等时ARMA模型的表达式.接着, 基于Kaimal风速谱,使用ARMA模型来模拟一座实际高架桥的脉动风速时程.最后,通过比较模拟风速功率谱、自相关和互相关函数与目标风速功率谱、自相关和互相关函数的吻合程度, 验证基于ARMA模型模拟高架桥脉动风速时程的可行性.     

基于组合算法的金属铣削毛刺预测

利用最优权值系数,将灰色理论、人工神经网络和遗传算法有机结合,构建组合算法,依据3种数学方法建立3种组合模型：组合算术平均模型、组合平方和平均模型以及组合比例平均模型,并分别将3种组合模型应用于45钢铣削毛刺的预测．利用3个预测误差评价指标,即平方和误差指标、平均绝对误差指标和平均相对误差指标,对各模型的预测结果进行分析计算.结果表明,组合算术平均模型所得结果与实验结果取得了较好的吻合,具有较高的精度和稳定性,对于金属铣削毛刺的预测具有实际的应用价值．     

Relative contributions of aleatory and epistemic uncertainty sources in time series prediction

This paper develops a novel computational framework to compute the Sobol indices that quantify the relative contributions of various uncertainty sources towards the system response prediction uncertainty. In the presence of both aleatory and epistemic uncertainty, two challenges are addressed in this paper for the model-based computation of the Sobol indices: due to data uncertainty, input distributions are not precisely known; and due to model uncertainty, the model output is uncertain even for a fixed realization of the input. An auxiliary variable method based on the probability integral transform is introduced to distinguish and represent each uncertainty source explicitly, whether aleatory or epistemic. The auxiliary variables facilitate building a deterministic relationship between the uncertainty sources and the output, which is needed in the Sobol indices computation. The proposed framework is developed for two types of model inputs: random variable input and time series input. A Bayesian autoregressive moving average (ARMA) approach is chosen to model the time series input due to its capability to represent both natural variability and epistemic uncertainty due to limited data. A novel controlled-seed computational technique based on pseudo-random number generation is proposed to efficiently represent the natural variability in the time series input. This controlled-seed method significantly accelerates the Sobol indices computation under time series input, and makes it computationally affordable.