基于改进杂草算法优化的神经网络模型在径流预报中的应用

针对BP神经网络在径流预报中易陷入局部最优解的缺陷及智能优化算法的优势,引入改进的杂草算法优化神经网络权值和阈值,将传统的杂草算法个体以正态分布空间扩散的方式改进为混合种群多种分布的方式产生子代个体。以金沙江流域中长期径流预报为例,将改进杂草算法优化的神经网络模型的径流预报结果与传统的BP神经网络和基于遗传算法优化的神经网络模型的预报结果进行对比。结果表明,改进杂草算法优化的神经网络应用到金沙江流域的径流预报精度较高,模型收敛更快,结果更加稳定,在实际预测中合理可行,具有一定的应用优势。研究成果为径流预报提供了新思路。     

基于改进BP网络的日径流预报模型研究

针对常规BP算法收敛速度慢和难以获得全局最优的不足,将网络误差函数的改变量引入权值和偏移值的调整,采用自适应学习速率和自适应动量因子调整策略,建立了基于多层感知器神经网络(MLP-ANN)的水文预报模型.采用自相关函教(ACF)和交又相关函数(CCF)确定网络输入因子并使用试错法优化网络结构.以湖南省双牌水库日入库流量预测为应用实例,并将模拟结果与常规BP网络模型和新安江模型进行对比分析.结果表明,改进模型收敛速度快、预报精度高.     

ABC BP模型在混凝土双曲拱坝变形监控中的应用

因大坝变形具有很强的非线性、随机性,致使预测困难,将人工蜂群算法(ABC)与BP神经网络相结合,利用人工蜂群算法具有强全局优化能力、强鲁棒性等优点,克服BP神经网络收敛速度慢、易陷入局部极小点等缺点,建立ABC BP、BP神经网络大坝变形预测模型对小湾大坝变形监测数据进行预测。结果表明,与单纯的BP神经网络预测模型相比,ABC BP算法提高了大坝变形预报的精度,加快了网络的收敛速度,能更高效准确地进行大坝变形监控预报。     

燃煤锅炉NO_x排放浓度的智能预报

在热态试验数据的基础上,分别应用BP(神经网络)和SVM(支持向量机)回归算法建立了燃煤机锅炉NOx排放特性模型,并验证了模型的准确性。结果表明,BP网络模型对检验样本的最大预测误差、最小预测误差和均方差分别为4.263%、0.556%和2.2133%,支持向量机模型对检验样本的最大预测误差、最小预测误差和均方差分别为2.121%、0.091%和0.4549%。两种智能技术都能对锅炉在不同工况下的NOx排放做出较为准确的预报,但支持向量机在泛化能力、收敛速度、最优性等方面明显优于神经网络。     

DGA SVR日径流非线性预报模型及应用

实时而准确的日径流量预报在防洪减灾、优化调度等方面起到了巨大作用。将遗传算法(GA)与支持向量回归(SVR)改进模型耦合,同时对SVR三个重要参数(C,σ,ε)进行动态寻优,构建了动态三参数优化GA-SVR日径流非线性预报模型(DGA-SVR)用于黑水河流域日径流预报,通过与BP神经网络和多元线性回归预测结果进行对比分析,DGA-SVR模型预测精度明显优于BP神经网络和多元回归模型。     

基于MPGA-LSTM月径流预测模型及应用

鉴于径流预测在水利基础设施运行、防洪、航运规划等水资源规划管理中具有极其重要的意义,将长短时记忆神经网络(LSTM)和多种群遗传算法(MPGA)相耦合,构建了基于MPGA-LSTM的混合径流预测模型,并采用该模型对晋江上游控制性水文站石砻站2013～2017年逐月径流过程进行延时回归预测。同时,为了验证MPGA-LSTM模型在月径流预测方面的优越性,针对影响预测精度三个因素基础模型、控制参数和优化方法,分别设置三组对比模型。计算结果表明,MPGA-LSTM模型表现最优,合格率为85%,确定性系数R~2为0.953,达到预报甲等标准。因此,所提出的MPGA-LSTM模型为月径流预测提供了一种有效的方法。     

基于PSO-WNN的大坝安全监控模型研究

针对大坝安全监控中小渡神经网络模型(WNN)训练时间较长且易陷入局部极小值的缺陷,提出采用粒子群优化算法(PSO)取代传统的梯度下降法对小波神经网络中的各参数进行优化.建立了PSO-WNN模型并用于大坝安全监测的拟舍和预报.实例结果表明,PSO-WNN模型收敛速度快、预测精度及稳定性高,为大坝变形监测分析提供了一种有效的新型建模方法.     

基于BP神经网络与遗传算法风电场超短期风速预测优化研究

风速预测对于风力发电并网调度至关重要。基于BP神经网络建立了风速预测模型,并从BP算法及遗传算法自身特点出发,针对BP网络结构确定困难、收敛速度慢等问题,提出创建多种群遗传算法,实现对BP神经网络的结构和权值初始值的同步优化。通过具体算例表明,经优化后的BP算法的收敛步数和计算时间明显减少,预测精度更高,网络整体性能有了显著提高。     

考虑预报因子选择的神经网络降雨径流模型

为优化神经网络模型的应用效果,研究了基于神经网络的降雨-径流模型,根据Copula熵法确定预报因子,并与传统的线性相关法进行比较分析,采用BP、RBF、GRNN三种神经网络建立降雨-径流模型,应用均方根误差、合格率、确定性系数三个指标为模型选取评价准则。通过对金沙江流域的径流预报,发现基于Copula熵法的BP模型预报结果更接近实测值,精度更高。     

基于万有引力搜索算法优化BP的尾矿坝浸润线预测

为解决BP神经网络在尾矿坝浸润线预测中易陷入局部最小值、收敛速度慢等问题,引入全局搜索能力强的万有引力搜索算法(GSA)优化BP神经网络的权值和阈值,进而构建万有引力搜索算法优化BP神经网络(GSA-BP)模型,并以西果园南峪沟尾矿坝NJRX2-4监测点的浸润线预测为例,对比分析GSA-BP模型的预测值与BP模型和PSO-BP模型的预测值。结果表明,GSA-BP模型构建合理,预测精度最高,预测值与实测值相近,验证了GSA-BP模型在西果园南峪沟尾矿坝浸润线预测中的可行性和有效性。     

Effect of corrosion inhibitors on hydrogen uptake by Al from NaOH solution

Corrosion rate, hydrogen permeation rate (hydrogen uptake) and stress corrosion cracking of Al were studied in NaOH solutions, pure and with the addition of H₃BO₃, EDTA, KMnO₄ and As₂O₃. The presence of the studied species in electrolyte and the implantation of Al surface with B⁺ ions inhibited corrosion. Hydrogen uptake was found to be promoted or inhibited by means of studied species, depending on the method of their introduction into the base solution and on the applied polarization. The observed different influence of corrosion inhibitors on the hydrogen uptake was associated with the different chemical composition and structure (revealed by XPS analysis) of the surface films, formed on Al under the various conditions. Under similar polarization conditions, the presence of H₃BO₃ in the base solution similarly affected the hydrogen uptake by Al and the susceptibility to stress corrosion cracking of the metal.     

C在Ni(111)表面吸附行为的密度泛函理论研究

应用密度泛函理论系统地研究C在Ni(111)表面上的吸附能、吸附结构、差分电荷密度、局域态密度及Mulliken布居数,给出了覆盖度在0.167~1.0 ML内,C的吸附特性随覆盖度变化的规律。研究表明,C的稳定吸附位为三重空位(hcp,fcc),化学吸附成键的本质是C2p态和Ni3d态的耦合、杂化,使原本孤立的C原子态耦合杂化为成键态和反键态。随着覆盖度的增加,反键态电子占据增多,成键态电子减少,平均成键电子数减少,C-Ni间的相互作用减弱,吸附能降低。     

Effect of carbon nanotube loadings on supercapacitor characteristics

In order to improve the properties of electrodes based on carbon nanotubes (CNTs), nanocomposite electrodes of CNTs/activated carbon (AC) are prepared. The effect of CNT loadings on the performance of the nanocomposite electrodes is evaluated by electrochemical methods. Scanning electron microscopy images show the CNTs are well dispersed, entwine with the AC, and form a more conductive network in a nanocomposite electrode. When CNT loadings increase from 0 to 10 wt%, the capacitance increases by 15% (89.4 to 102.5 F g^?1), and the equivalent series resistance (ESR) decreases by 13% (0.93 to 0.81 Ω). The capacitance improvement has an optimum CNT loading of about 6 to 10 wt%. The decreased ESR with high CNT loadings does not prevent capacitance loss at sweep rates greater than 50 mV s^?1. The nanocomposite electrode shows good cycle performance with no capacity loss after 5000 cycles at a scan rate of 500 mV s^?1. These results indicate that the CNT‐based nanocomposite electrodes would be a promising material for use in supercapacitors. Copyright © 2014 John Wiley & Sons, Ltd.     

废气再循环对二甲醚/甲醇均质压燃燃烧过程影响的试验研究

在单缸柴油机上进行了冷却废气再循环(EGR)对二甲醚(DME)/甲醇均质压燃(HCCI)燃烧过程影响的试验研究。结果表明,EGR对拓宽二甲醚/甲醇HCCI发动机的最大负荷作用不大;随着EGR率增大,主燃烧开始时刻和放热峰值明显后移,主燃烧持续期延长,放热峰值降低。EGR率为25%时的最大爆发压力比没有EGR时降低了近1.3 MPa,最大爆发压力出现的位置推迟了7°CA;EGR率增大,二甲醚/甲醇HCCI发动机的指示热效率升高。对应给定的EGR率,存在一个热效率较高的DME比例区间;HC和CO排放随EGR率的增大而增加,随DME比例的增加而降低,NOx排放接近于零。控制EGR率和DME比例是控制二甲醚/甲醇HCCI发动机燃烧过程、性能和排放的关键。     

Whisker formation on a thin film tin lithium-ion battery anode

Tin (Sn) is a candidate material for anodes (negative electrodes) of lithium-ion batteries because of its high theoretical energy capacity. In this paper, we report an observation of Sn-whisker growth on Sn-thin films after lithiation and delithiation. The compressive stress generated by electrochemical lithiation of the Sn-thin films is likely the driving force for the growth of the Sn whiskers. Attention should therefore be paid to the issue of Sn-whisker growth for Sn-based electrodes since Sn whiskers may penetrate through the separator, and short-circuit the electrochemical cell.     

废气再循环对二甲基醚均质压燃燃烧过程影响的试验研究

在一台单缸发动机上进行了废气再循环(EGR)对二甲基醚(DME)均质压燃(HCCI)燃烧过程影响的试验研究。结果表明，EGR比例小于20％对运行最大负荷工况范围影响不大；采用高比例EGR可以拓宽DME均质压燃运行工况范围，随着EGR率增大，HCCI运行的最大负荷工况增大，着火燃烧时刻推迟，燃烧放热率降低，缸内最大爆发压力降低，发动机热效率增大；EGR率小于75％，HC排放略有降低或相当，EGR率为75％时，HC排放显著增加；EGR率大于25％，随着EGR率增加，CO排放增大，小负荷工况尤其明显，在中高负荷工况，EGR率对CO排放影响较小。     

一种基于变换器拓扑的新型MPPT控制算法的研究

提出一种基于变换器拓扑（Boost／Buck）的最大功率点跟踪控制方法,该方法既能准确、稳定地跟踪到最大功率点,又只需对光伏输出电压这一个模拟量进行采样。同时,对所提出的该种方法进行了仿真建模分析,并比较了这几种MPPT算法下的跟踪效果。结果表明该方法具有更好的快速性、稳定性和经济性。     

废气再循环与燃料辛烷值对均质压燃发动机性能和排放影响的试验研究

在一台单缸直喷式柴油机上研究了废气再循环（EGR）对不同辛烷值燃料均质压燃（HCCI）发动机性能和排放特性的影响。结果表明，混合气较稀，EGR对指示热效率影响较小，其影响和燃料辛烷值有关；混合气变浓，EGR对指示热效率的影响增大。不同辛烷值燃料最高指示热效率出现在高EGR率、混合气较浓的区域，并且靠近爆震燃烧边界，辛烷值为60的燃料最高指示热效率最高，并且覆盖的工况区域最宽。高EGR率区域，EGR对HC排放的影响十分明显，EGR率升高，HC排放急剧增大，而且随着燃料辛烷值增大，这种趋势越明显；CO排放与缸内燃烧温度有较大的相关性，EGR率升高，CO排放升高。NOx排放出现急剧升高的“拐点”是判断HCCI爆震燃烧的一个重要判据，EGR率增大，“拐点”出现的混合气浓度增大，在正常工作范围内，NOx排放极低，EGR对NOx排放几乎没有影响。     

Efficient hydrogen evolution catalysis triggered by electrochemically anchored platinum nano-islands on functionalized-MWCNT

We report the electrochemical deposition (ECD) of platinum nano-islands (Pt NIs) on functionalized multi-walled carbon nanotubes (ECD Pt NIs@f-MWCNT) as an efficient electrocatalyst for the hydrogen evolution reaction (HER). Pristine MWCNT was acid treated to induce the number of oxygen functional groups on the surface and enhances the wettability. Thereafter, Pt nanoparticles (Pt Nps) were deposited by a simple electrodeposition technique on the oxygen enriched MWCNT surface. The Pt NIs@f-MWCNT has been physicochemically characterized using X-ray diffraction (XRD), high resolution transmission electron microscopy (HRTEM), Raman spectroscopy and X-ray photoelectron Spectroscopy (XPS). The TEM analysis showed the presence of Pt NIs on MWCNT wherein, the NIs were made up of small Pt nanoclusters of ～4 nm in dimension. The electrochemical HER studies were carried out using linear sweep voltammetry (LSV), Tafel polarization and electrochemical impedance spectroscopy (EIS). An overpotential (?) of ?84 mV was obtained at a current density (j) of ?10 mA/cm². The amount of Pt loading has been optimized through electrodeposition. Enhanced HER activity was observed with a Pt loading of 3.8 μg/cm². In order to ascertain the durability of the catalyst, accelerated degradation test (ADT) was carried out for 10,000 cycles at a scan rate (?) of 100 mV/s. The turnover frequency (TOF) was estimated to be 6.3 s^?1 at ? = ?70 mV.     

Cracked palladium films on an elastomeric substrate for use as hydrogen sensors

We have investigated a lithography-free technique for On-Off type hydrogen sensors using a cracked palladium (Pd) film on an elastomeric substrate. Cracks were induced in a sputtered Pd film simply by undergoing hydrogen absorption and desorption processes. Compared to the same thickness of a Pd film on a Si/SiO₂ substrate that relied on the electron scattering mechanism, a cracked Pd film on an elastomeric substrate operated as a reversible On-Off hydrogen sensor based on the crack open-close mechanism when exposed to hydrogen. The thickness of a Pd film on the elastomeric substrate plays a significant role in determining the sensing mode of the cracked Pd film. The cracked Pd film with a thickness of 9–11 nm on the elastomeric substrate showed reversible and perfect On-Off responses under a wide range of hydrogen concentrations with large current variations and a fast response time of less than 1 s.