一种参数自调整风电功率预测模型

为了提高风电功率预测精度,提出一种参数自调整风电功率预测模型.通过加权递推最小二乘(SWWRLS)方法建立预测模型,侧重当前数据对预测结果的影响,排除了历史数据对预测结果的干扰.模型通过加权递推的方法节省了存储空间,并且提高了模型对外界环境数据变化的自适应性.最后,分别采用支持向量机(SVM)方法、卡尔曼滤波(KF)方法和本文SWWRLS方法,以辽宁省某风电场的真实历史数据进行风电功率预测对比实验,实验结果表明,本文方法建立的模型具有较高的预测精度.     

基于累积运动能量图像的人体动作识别

指出了动作识别中的最大困难是难以提取有效的特征来准确描述人体的动作,动作模板是众多方法中的一种简单有效的方法,用来描述动作特征的经典动作模板是运动历史图像.由于受噪声的干扰,用运动历史图像描述复杂环境下的人体动作并不十分理想.为了得到比运动历史图像更加有效的动作模板,提出了将视频序列中的运动能量信息用一张图描述出来,称之为累积运动能量图像,提取其直方图特征来表征人体动作.经You Tube数据集上的实验表明:该累积运动能量图像的识别率比同类方法高.     

New low-cost tubular ceramic microfiltration membrane based on natural sand for tangential urban wastewater treatment

In wastewater treatment, the development of low-cost separation methods is of significant importance. Low-cost membranes based on natural materials have become a highly active research topic in recent years. Herein, using low-cost natural Moroccan sand, new ceramic supports have been developed and characterized using different techniques such as X-ray diffraction (XRD), Fourier transform infrared (FTIR) spectroscopy, thermogravimetric analysis (TGA), differential thermal analysis (DTA), along with scanning electron microscope (SEM). Plastic paste (average particle size ≤125 µm) was blended with organic additives and water, then the obtained paste was extruded into porous tubular supports. The support had a porosity of 43%, water permeability of 1928 L/h m² bar, excellent chemical and mechanical properties and an average pore diameter in the range of 8–15 µm after firing at 950 °C/2 h. As per SEM analysis, the tubular supports had a smooth and crack-free surface. The slip casting process was used to create a microfiltration layer from the same natural sand powder (average particle size ≤63 µm) using a mixture of powder sand, water, and polyvinyl alcohol solution. The water permeability of the microfiltration membrane sintered at 950 °C/2 h was 1052 L/h m² bar, the average pore size diameter was about 0.90 µm and 82% of pores had a diameter ≤1.00 µm. The obtained microfiltration membrane was tested for the treatment of urban wastewater. The membrane showed excellent separation performance in turbidity removal and chemical oxygen demand.     

Composite Polybenzimidazole Membrane with High Capacity Retention for Vanadium Redox Flow Batteries

Currently, energy storage technologies are becoming essential in the transition of replacing fossil fuels with more renewable electricity production means. Among storage technologies, redox flow batteries (RFBs) can represent a valid option due to their unique characteristic of decoupling energy storage from power output. To push RFBs further into the market, it is essential to include low-cost materials such as new generation membranes with low ohmic resistance, high transport selectivity, and long durability. This work proposes a composite membrane for vanadium RFBs and a method of preparation. The membrane was prepared starting from two polymers, meta-polybenzimidazole (6 μm) and porous polypropylene (30 μm), through a gluing approach by hot-pressing. In a vanadium RFB, the composite membrane exhibited a high energy efficiency (~84%) and discharge capacity (~90%) with a 99% capacity retention over 90 cycles at 120 mA·cm⁻², exceeding commercial Nafion^® NR212 (~82% efficiency, capacity drop from 90% to 40%) and Fumasep^® FAP-450 (~76% efficiency, capacity drop from 80 to 65%).     

Ultrasonically induced ZnO–biosilica nanocomposite for degradation of a textile dye in aqueous phase

In the present study, a porous clay-like support with unique characteristics was used for the synthesis and immobilization of ZnO nanostructures to be used as sonocatalyst for the sonocatalytic decolorization of methylene blue (MB) dye in the aqueous phase. As a result, the sonocatalytic activity of ZnO–biosilica nanocomposite (77.8%) was higher than that of pure ZnO nanostructures (53.6%). Increasing the initial pH from 3 to 10 led to increasing the color removal from 41.8% to 88.2%, respectively. Increasing the sonocatalyst dosage from 0.5 to 2.5 g/L resulted in increasing the color removal, while further increase up to 3 g/L caused an obvious drop in the color removal. The sonocatalysis of MB dye over ZnO–biosilica nanocomposite was temperature-dependent. The presence of methanol produced the most adverse effect on the sonocatalysis of MB dye. The addition of chloride and carbonate ions had a negligible effect on the sonocatalysis, while the addition of persulfate ion led to increasing the color removal from 77.8% to 99.4% during 90 min. The reusability test exhibited a 15% drop in the color removal (%) within three consecutive experimental runs. A mineralization efficiency of 63.2% was obtained within 4 h.     

高应力软岩下矿井巷道支护

随着煤矿开采深度的不断增加,井下煤矿巷道将处于更高的地应力环境中,尤其在地质构造活动强烈的地区,井下巷道支护及稳定性更加难以保证.研究高地应力软岩环境下巷道科学的支护方式是保证煤矿采掘深部煤层的关键.就高应力软岩的基本概念及形成条件进行了讨论,在掌握高应力软岩巷道的变形特征和支护原理的基础上,以芙蓉矿区白皎煤矿井下高应力软岩巷道支护为对象,对3种不同的支护方式进行了研究与实践,结果表明预留刚隙柔层支护方式为最佳的适合高应力软岩条件下的巷道支护方式.     

综放工作面顶煤承载能力和支架工作阻力的探讨

支架工作阻力计算是综放开采研究的主要问题之一.在对顶煤力学性质进行了实验室试验的基础上,总结出顶煤破坏时的强度满足主应力表示的莫尔-库仑强度准则.根据综放工作面顶煤体的受力特点,运用损伤力学应变等效原理,建立了综放工作面控顶距内顶煤体受力力学模型.从顶煤体承载能力的角度,研究综放面支架-顶煤体之间相互作用关系,推出了支架工作阻力的理论计算公式.该模型较好地解释了综放工作面支架工作阻力随顶煤强度的提高而增加的矿山压力显现.研究结论对综放面支架工作阻力计算具有一定的理论指导意义.     

基于SVM的置信度综合方法在语音识别中的应用

讨论了语音识别中使用支持向量机(support vector machines,SVM)对音子级置信度进行综合的方法. 音子级置信度得分采用传统的方法计算而得,并使用SVM对音子级置信度进行综合得到词级的置信度得分. 在说话人无关的汉语孤立词识别实验中,使用作者方法比使用传统方法获得的系统等错误率rEER(equal error rates,EER)有明显降低,可以从基线系统的28.14%降低到23.71%,而系统的复杂度仅有小幅度的上升.     

基于网络的模型库结构表示方法研究

针对现有方法存在的不能表示复杂模型关系以及模型组合与求解自动化程度低的局限性,在对模型库结构表示方式进行比较的基础上,从而进一步分析并提出了数据类型广义匹配的概念,将其应用于模型连接的表示,给出了模型库的网络结构表示的定义及其构造方法。分析了模型组合及模型库必须考虑的元素,并用网络结构表示模型库中元素间的静态连接结构;给出了模型库网络结构的形式化定义,模型间的连接通过引入类型转换实现。从而简化模型之间连接,使模型的连接判断由类型匹配判断转化为模型与转换和转换与模型之间的类型相等性判别,而类型匹配操作则通过转换自动完成。并且在组合模型求解过程中,转换过程也会自动完成,提高了模型组合与求解的自动化。     

基于RS与LS-SVR的储层参数预测

该文提出结合粗糙集(RS)和最小二乘支持向量回归(LS-SVR)的方法.该方法利用RS对原始数据进行约简,更好地减少了支持向量的维数;同时采用LS-SVR解决了常规SVM计算速度慢、抗噪能力差的缺点.实例证明该方法应用在复杂地层储层参数预测中具有优越性.