Electricity market price spike analysis by a hybrid data model and feature selection technique

In a competitive electricity market, energy price forecasting is an important activity for both suppliers and consumers. For this reason, many techniques have been proposed to predict electricity market prices in the recent years. However, electricity price is a complex volatile signal owning many spikes. Most of electricity price forecast techniques focus on the normal price prediction, while price spike forecast is a different and more complex prediction process. Price spike forecasting has two main aspects: prediction of price spike occurrence and value. In this paper, a novel technique for price spike occurrence prediction is presented composed of a new hybrid data model, a novel feature selection technique and an efficient forecast engine. The hybrid data model includes both wavelet and time domain variables as well as calendar indicators, comprising a large candidate input set. The set is refined by the proposed feature selection technique evaluating both relevancy and redundancy of the candidate inputs. The forecast engine is a probabilistic neural network, which are fed by the selected candidate inputs of the feature selection technique and predict price spike occurrence. The efficiency of the whole proposed method for price spike occurrence forecasting is evaluated by means of real data from the Queensland and PJM electricity markets.     

Wavelet-based feature extraction and selection for classification of power system disturbances using support vector machines

This paper presents a new approach for the classification of the power system disturbances using support vector machines (SVMs). The proposed approach is carried out at three serial stages. Firstly, the features to be form the SVM classifier are obtained by using the wavelet transform and a few different feature extraction techniques. Secondly, the features exposing the best classification accuracy of these features are selected by a feature selection technique called as sequential forward selection. Thirdly, the best appropriate input vector for SVM classifier is rummaged. The input vector is started with the first best feature and incrementally added the chosen features. After the addition of each feature, the performance of the SVM is evaluated. The kernel and penalty parameters of the SVM are determined by cross-validation. The parameter set that gives the smallest misclassification error is retained. Finally, both the noisy and noiseless signals are applied to the classifier given above stages. Experimental results indicate that the proposed classifier is robust and has more high classification accuracy with regard to the other approaches in the literature for this problem.     

Four-node mixed Hu–Washizu shell element with drilling rotation

In this paper, enhanced four-node shell elements with six DOFs/node based on the Hu–Washizu (HW) functional are developed for Green strain. The drilling rotation is included through the drilling rotation constraint equation. The key features of the approach are as follows.

1. The shell HW functional is derived from the shell potential energy functional, which is an alternative to the derivation from the three-dimensional HW functional. This method is more versatile as it enables the derivation of the so-called partial HW functionals, with different treatment of the bending/twisting part and the transverse shear part of strain energy.
2. For the membrane part of HW shell elements, a seven-parameter stress, a nine-parameter strain and a two-parameter enhanced assumed displacement gradient enhancement are selected as optimal. The assumed representations of stress and strain are defined in skew coordinates in the natural basis at the element's center. This improves accuracy and has positive theoretical consequences.
3. The drilling rotation constraint equation is treated by the perturbed Lagrange method. The faulty term resulting from the equal-order approximations of displacements and the drilling rotation is eliminated, and one spurious mode is stabilized using the gamma method. The proposed formulation is insensitive to the element's distortions and yields a large radius of convergence in the examples involving in-plane bending.

The performance of 4 four-node shell HW elements, having different bending/twisting and transverse shear parts, is analyzed on several numerical examples. Such aspects are considered as: accuracy, radius of convergence, required number of iterations of the Newton method or the arc-length method and time of computations. The element with 29 parameters (HW29) is selected as the best performer. Copyright © 2011 John Wiley & Sons, Ltd.     

An automatic impact-based delamination detection system for concrete bridge decks

Delamination of concrete bridge decks is a commonly observed distress in corrosive environments. In traditional acoustic inspection methods, delamination is assessed by the “hollowness” of the sound created by impacting the bridge deck with a hammer or bar or by dragging a chain. The signals from such sounding methods are often contaminated by ambient traffic noise and delamination detection is highly subjective. In the proposed method, a modified version of independent component analysis (ICA) is used to filter the traffic noise. To eliminate subjectivity, mel-frequency cepstral coefficients (MFCC) are used as features for delamination detection and the delamination is detected by a radial basis function (RBF) neural network. Results from both laboratory and field data suggest that the proposed method is noise robust and has satisfactory performance. The method can also detect the debonding of repair patches and concrete delamination below the repair patches. The algorithms were incorporated into an automatic impact-based delamination detection (AIDD) system for field application.     

Processing of nested and cross-serial dependencies: an automaton perspective on SRN behaviour

Language processing involves the identification and establishment of both nested (stack-like) and cross-serial (queue-like) dependencies. This paper analyses the behaviour of simple recurrent networks (SRNs) trained to handle these types of dependency individually and simultaneously. We provide new converging evidence that SRNs store sequences in a fractal data structure similar to a binary expansion. We provide evidence that the process of recalling a stored string by an SRN depletes the stored data structure, much like the operations of a symbolic stack or queue. Trained networks do not seem to operate like random access arrays, where a pointer into a data structure can retrieve data without altering the contents of the data structure. In addition, we demonstrate that networks trained to model both types of dependencies do not implement a more complex, but unified, representation, but rather implement two independent data structures, similar to a stack and queue.     

基于MSER的红外与可见光图像关联特征提取算法

关联特征提取是红外与可见光图像配准、融合和变化检测等应用中的关键步骤。针对同一场景红外与可见光图像间关联特征难以正确提取的问题,基于最稳定极值区域(MSER)算法,提出了一个仿射不变的关联特征提取方法。该方法主要包括3个步骤:(1)提取红外与可见光图像中的最稳定极值区域;(2)对特征区域进行椭圆拟合;(3)规则化处理,消除形变干扰,输出便于描述和匹配的一致性特征。实验结果证明了该算法在红外与可见光图像关联特征提取中的有效性。     

遥感影像检索中高维特征的快速匹配

提高特征点匹配效率是将高维局部特征运用于遥感影像检索的关键,该文提出一种新的压缩优先过滤(CPF)索引算法。该算法通过量化特征向量构建近似向量空间上的高维索引结构,利用优先队列过滤得到近似近邻候选集,精确计算候选实际特征向量得到最终近邻。在CPF算法基础上提出了基于快速鲁棒性特征(SURF)的遥感影像快速检索算法。实验及分析表明,与经典的最佳桶优先(BBF)算法相比较,CPF降低了磁盘读写(I/O)和浮点运算次数,特征点数目较大时,查询效率和总体查询精度均有显著提高,基于SURF特征的遥感影像快速检索算法能快速返回正确目标与相似目标影像。     

云南昭通炎山小田村铅锌矿床成矿控制条件分析

矿区位于滇黔川铅锌成矿区之巧家-金沙厂成矿带之上,处于金沙厂、茂租-东坪两大矿化集中区之间。区内震旦系灯影组上部硅质条带白云岩、顶部含藻白云岩为矿区赋矿地层,矿体呈似层状、透镜状及沿断裂带呈脉状产于以上地层中,其控矿因素为地层-岩性(相)-构造。找矿标志为灯影组上部、顶部的硅质条带白云岩、含藻白云岩,莲峰-茂租区域控矿断裂上盘和旁侧的次级断裂,相应的围岩蚀变发育的地区(段)。     

基于小世界模型的流形学习算法

等距特征映射(ISOMAP)不仅计算复杂度很高,而且缺乏对新样本的学习能力。基于标志点的ISOMAP(L-ISOMAP)通过只保持一些标志点之间的测地线距离有效地降低了复杂度,然而标志点集的随机选择常常会导致较差的嵌入结果。为此,提出了一种基于小世界模型的流形学习算法。根据小世界模型的原理,该算法仅仅保持每个样本点与其k个最近邻和一些随机选择的远点之间的测地线距离,采用最速梯度下降法优化来得到数据的低维表示。理论分析表明,该算法的计算复杂度远远低于ISOMAP的复杂度。利用应力函数和剩余方差对3个算法进行了比较。实验结果表明,从该算法得到的结果与从ISOMAP得到的结果相近,且优于从L-ISOMAP得到的结果。同时,该算法可以实现对新样本的学习,对噪声也不太敏感。     

基于双视线特征感知编组的影像匹配方法

与点特征相比,高级别的特征,特别是自由形状线特征包含了更加丰富的信息,这对于其后的影像解译和表面重建等处理是更为有利的。针对传统自由形状线特征匹配方法的弊端——匹配成功的线特征数量较少,引入了感知编组理论,提出了一种基于双视线特征感知编组的影像匹配方法。实验证明该方法不仅可以提高成功匹配的自由形状线特征数量,还可增加自由形状线特征的长度,方便后续人工地物（如建筑物）的提取。