GPS数据驱动的车辆跟驰行为建模与分析

为研究道路交通流特性,基于车载高精度GPS跟驰试验数据进行车辆跟驰建模研究,结合深度学习理论和数据驱动方法,构建了基于粒子群优化(particle swarm optimization, PSO)的长短期记忆(long short term memory, LSTM)车辆跟驰模型。首先,清洗和平滑车辆轨迹数据,并对驾驶特征行为参数及相关关系进行研究,如加速度、车头时距以及速度与跟驰距离特性关系等。在此基础上,制定跟驰状态筛选规则;其次,构建考虑时间序列的PSO-LSTM模型,识别跟驰数据样本集,将当前时刻的前车速度、车头间距和上一时刻的车头时距作为模型输入,预测当前时刻的跟驰车速度;接着,选用25辆车跟驰试验的高精度GPS数据验证PSO-LSTM模型性能;最后,为验证该模型的优越性,选用传统机器学习SVR(support vector regression)模型以及深度学习LSTM模型作为对比。结果表明,基于粒子群优化的长短期记忆模型预测精度高达0.993,整体预测效果高于SVR模型和LSTM模型,其中预测误差指标MAPE(mean absolute percentage error)较SVR和LSTM分别降低了60.02%、1.52%。PSO算法进行超参数优化后的PSO-LSTM模型,能更好地模拟车辆的跟驰行为。     

基于粒子群布谷鸟融合算法的主汽温系统控制器参数优化

火电厂主汽温系统具有大惯性、大时滞特性,经典串级比例-积分-微分控制器(PID)难以实现对主汽温系统的精细控制,本文利用粒子群与布谷鸟的融合算法对主汽温控制系统的控制器参数进行优化,通过选择合适的目标函数,对某600 MW直流锅炉主汽温控制系统进行了优化,与工程整定法、粒子群算法的控制结果相比,粒子群布谷鸟融合算法具有...     

Four new species of Hargicotyle Mamaev, 1972 (Diclidophoridae) parasites on sciaenid fishes from Peru and Chile

Hargicotyle magna sp. nov., Hargicotyle paralonchuri sp. nov., Hargicotyle sciaenae sp. nov., from the gills of the Sciaenids Sciaena fasciata, Paralonchurus peruanus and Sciaena deliciosa, respectively, and Hargicotyle menticirrhi sp. nov. from the gill and mouth of Menticirrhus ophicephalus are described from the northern Chilean and central Peruvian coast. Distinct characteristics of the new species are the distribution of the vitelline follicles; the number and distribution of the testes, the size, shape and number of larval hooks; and the presence of one or two suckers in each clamp.     

Metazoan parasites of Calophysus macropterus (Siluriformes: Pimelodidae) in the Acre and Iaco rivers in the western Amazon region of Brazil: diversity,similarity and seasonal variation

ABSTRACT

The diversity, similarity and seasonal variation of metazoan parasite communities in Calophysus macropterus in the Acre and Iaco rivers, in the western Amazon (Brazil), was investigated. Parasites from 13 taxa were collected from C. macropterus in both rivers: four species of monogeneans, four nematodes, two cestodes, one digenean, one crustacean and one pentastomid. In hosts from the Acre river, Cucullanus pinnai predominated; while in hosts in the Iaco river, Monticellia amazonica predominated. The component communities of the parasites among the hosts in the two rivers presented high similarity (100%). Prevalence of Alinema amazonicum was higher in hosts in the Acre river; while the prevalence of C. pinnai was higher in hosts in the Iaco river and the mean prevalence and abundance of M. amazonica were higher in fish from the Iaco river. Regarding C. macropterus from the Acre river, infection levels by A. amazonicum were higher during the rainy season, while Demidospermus pinirampi only occurred in the dry season and Procamallanus (Spirocamallanus) inopinatus only occurred in the rainy season. In hosts from the Iaco river, infections by larvae of Anisakidae gen. sp. were higher during the dry season, while infection by Rudolphiella piracatinga and Sebekia sp. only occurred in the rainy season. However, P. (S.) inopinatus, Ergasilus callophysus, Ameloblastella unapi, Demidospermus luckyi, Demidospermus macropteri and D. pinirampi only occurred in the dry season. High similarity of the component communities of the parasite was observed between the rainy and dry seasons. These results suggest that factors other than location and seasonality were influencing the communities and infracommunities of the parasites found. Lastly, C. macropterus is a new host for almost 50% of the parasite species found. In addition, the results from the present study have expanded the geographical range of these 13 species of parasites to the western Amazon region.     

电解法处理鱼粉废水

研究实验室条件下电解法对模拟鱼粉废水处理的可行性.通过单因素实验和正交实验确定了在pH=6、电解质投加0.6 g、直流电压20 V、电解40 min条件下化学耗氧量(COD)去除率达到88.4%,鱼粉废水经过电解处理后COD达到《辽宁省沿海地区污水直排入海域标准》,能够取得较好的有机物去除效果,具有工业应用前景.     

人工鱼高级行为规划和运动模型研究

以生物学理论为依据,从人工智能和人工生命相结合的角度提出了人工鱼的高级行为规划和运动模型.高级行为规划即设计多种行为序列以实现当前意图,使人工鱼能够根据不同的环境条件选择相应的行为策略.运动模型在高级行为规划指导下,将行为序列分解为具体动作来执行,作用于环境.人工鱼的运动由饥饿度、危险度等内部状态驱动,并受体力的影响.仿真实验表明该模型能够更合理、有效地模拟真实鱼类的生命特征和智能行为.     

基于动能PSO的环己醇/环己酮动力学参数估计

针对基本PSO算法在迭代后期粒子发生"趋同"而易陷入局部极值的问题,提出了动能粒子群算法(KEPSO)。该算法将粒子"趋同"看作粒子群体与最优粒子发生塑性碰撞的过程,通过动能补偿机制使"惰性"粒子重新恢复"活力",从而跳出局部极值。仿真结果显示,KEPSO算法大大提高了全局搜索能力,在高维函数测试中表现出了较好的优化性能。将KEPSO算法用于环乙醇/环已酮硝酸氧化动力学参数估计中,获得模型的平均相对误差绝对值之和比文献报道值分别降低了42.6%和47.3%。     

基于属性相似度的云分类器

提出了属性相似度概念解决高维对象分类的权重问题,并结合云理论建立了基于属性相似度的云分类器.采用云理论建立训练集的各属性模型,表达各属性值隶属于其类别中心Ex的程度.分类模型由属性模型集成得到,属性权重根据属性相似度计算.各类别的同一属性间的相似度越大,此属性对分类的作用越小.基于粒子群优化方法对分类模型的中心位置Ex进行优化.将此分类器与普通云分类器应用于iris数据集的分类实验,该分类器的分类效果好于后者.     

基于QoS-PSO的无线传感器网络路由方法

过去,大部分的无线传感器网络(WSN)的应用局限于数据的采集,比较少地关注网络中各节点的协同合作,现有的WSN路由算法对于网络的动态性的支持力度非常弱,不能满足分布式的无线网络和无线传感器网络应用对于网络服务质量的需求,路由优化能力不足.通过综合考虑多个服务质量指标,然后基于智能遗传算法微粒群算法进行路由寻优,为路由寻优提供了很好的搜索能力.仿真结果表明,基于微粒群优化算法,以综合服务质量(QoS)指标为目标的路由很好地改善了其服务质量性能.     

基于量子粒子群优化算法的Web服务组合问题

使用量子粒子群优化算法（QPSO）,将可能的web服务工作流执行路径看作粒子,按照QPSO算法进行进化,从而解决了基于服务质量（QualityofService,qos）~束的Web服务组合问题,此为解决Web服务组合问题提出了一种新的思路．实验表明,使用QPSO算法求解复杂web服务组合问题在组合时间上具有一定的优越性．