Evaluation of the Living Conditions and Health Perceptions of Waste Electrical and Electronic Equipment Handlers in the Informal Sector in Dakar

Electrical and electronic devices are becoming an increasingly important part of our society. In Africa, and in Senegal in particular, the handling and management of electronic and electrical equipment (EEE) that has reached the end of its life is mainly informal. This professional environment is characterized by the disintegration of the sector and the social heterogeneity that can be found there. The objective of this study is to assess the standard of living of electrical and electronic equipment waste handlers in the Dakar region, as well as their perception of their health. A survey was used to obtain information on sociodemographic background, living arrangements, perception of health status, and good practices to be implemented in case of work-related health problems. Life style, perception of general health and health problems were ranged as excellent, very good, good, average and poor. Informal recyclers in the Dakar region lived mainly in rooms and buildings as tenants (49.1%), or in family homes (48.4%) before starting this activity, and 51.2% continue to live in rooms and buildings as tenants compared to 41.4% who still live in a family home. The perception of health status ranged from poor to excellent, and 4.9% believe that they are limited in work due to a disability or health problem. Informal work is a heterogeneous phenomenon that makes research and policymaking particularly complex. There are several external factors within informal WEEE re-cyclers that can cause health problems or functional disability. However, the living conditions and the perception they have of their state of health are contradictory to the working conditions and the social environment to which they belong. A biomedical approach would consolidate these achievements by confirming or invalidating them.

高速铁路日常客运量的EMD-Informer组合预测方法

铁路客流需求的科学预测是进行运输组织方案决策的重要依据。以高速铁路历史客票数据为基础,结合经验模态分解(empirical mode decomposition,EMD)与机器深度学习中的注意力机制,提出高速铁路日客流量的EMD-Informer组合预测方法。首先采用EMD方法分解高速铁路客流量序列,获得具有周期特征和线路客流内在特征的模态分量,再利用Informer模型分别训练和预测各模态分解分量,并通过多头注意力机制高效挖掘客流数据内在规律和捕捉数据序列中的关键特征,在此基础上重组各分量预测值,从而得到高速铁路日常客流的整体高精度预测值。同时,根据结合问题特征的大量实验,明确可供实际运用参考的超参数设置规则。基于京沪高速铁路全线的实例计算分析表明,相对对比预测方法,EMD-Informer组合预测方法在高速铁路客流的单步预测及超前预测上均具有明显更小的预测误差。

基于STL-Informer-BiLSTM-XGB模型的供热负荷预测

供热负荷预测是指导供热系统调控的重要手段。提高供热负荷预测精度十分重要,针对机器学习中输出目标的分解预测,提出了一种基于季节和趋势分解(seasonal and trend decomposition using loess,STL)的供热负荷预测方法,构建了适用于供热负荷预测的输出目标。首先利用STL算法将供热负荷时间序列数据分解为趋势分量、周期分量和残差分量,分别训练Informer、BiLSTM和XGB模型,将构建好的3个分量预测模型的输出叠加作为初步预测结果,分析误差序列,以BiLSTM预测误差提高模型精度,构建出STL-Informer-BiLSTM-XGB预测模型。将上述模型与常用预测模型进行对比,结果表明所构建的STL-Informer-BiLSTM-XGB模型的MAPE、MAE和MSE分别为0.871%、96.18和13202.2,预测效果最优,验证了所提出的方法具有较高的供热负荷预测精度。

TimeGAN-Informer长时机场能见度预测

能见度的预测对机场的业务决策、保障飞机的安全起降具有重要的意义。针对现有能见度预测模型预测时间较短的问题,提出一种基于TimeGAN Informer(Time Generative Adversarial Network-Informer)的机场能见度预测方法。利用2018—2022年气象和污染物数据,通过相关系数法和递归特征消除法提取出能见度的主要影响因素,使用TimeGAN时间序列生成对抗网络对数据进行扩充,并将Informer长时间序列预测模型应用于能见度预测。结果显示:当预测步长为1 d、2 d、3 d时,TimeGAN Informer的绝对误差(Mean Absolute Error,MAE)分别为2.42、3.13、3.57,比Informer分别降低了0.29、0.27、0.28,比长短时记忆网络(Long Short-Term Memory,LSTM)分别降低了0.28、0.49、0.63;均方根误差(Root Mean Square Error,RMSE)分别为3.03、3.7、4.09,比Informer分别降低了0.38、0.22、0.24,比长短时记忆网络(LSTM)分别降低了0.3、0.5、1.04;百分误差小于30%的分别占测试样本集的78.07%、70.68%、63.84%。尽管随着步长的增加预测效果变差,但在预测步长为3 d时,多数样本的预测误差仍小于30%,实现了对机场区域较为准确的长时能见度预测。

基于KCR-Informer的长期风电功率预测研究

准确的长期风电功率预测对电网系统稳定运行至关重要,传统预测方法在处理长序列预测时效果并不理想,近期研究表明Informer模型在长序列预测领域取得良好效果。然而,该模型在捕捉数据的局部特征以及处理网络层数堆叠问题上还有待改进。文章提出一种基于卡尔曼滤波器-卷积神经网络-残差网络-Informer(Kalman filter-convolutional neural network-residual network-informer,KCR-Informer)模型的长期风电功率预测方法,首先分析气象数据对风电功率的影响,使用卡尔曼滤波器对风电气象数据进行数据平滑处理,以减轻噪声对数据的影响,然后基于Informer模型建立风电功率预测模型,根据气象数据以及历史功率数据进行长期功率预测;在此基础上,引入卷积神经网络和残差连接模块,使模型能够更好的捕捉到局部特征,同时加快模型收敛,解决模型网络退化问题。算例的结果表明,与长短期记忆网络(long short-term memory,LSTM)算法、Transformer算法、Informer算法相比,文章方法在不同预测步长下的平均绝对误差(mean absolute error,MAE)降低5.7%~30%,均方误差(mean square error,MSE)降低8.3%~35%,长期风功率预测的精度得到提升,验证了模型的有效性。

复杂环境下基于改进Informed RRT*的无人机路径规划算法

针对无人机在复杂环境中进行路径规划时,快速搜索随机树(RRT)算法易出现规划时间长、路径冗余、狭窄空间中易陷入局部约束导致规划失败的问题,提出一种改进的Informed RRT*算法.首先,引入人工势场法使采样点按照势场下降的方式向目标点移动,以提高RRT树扩展的目的性和方向性.然后,考虑随机树在扩展过程中全局环境的复杂度,引入自适应步长调整策略以增加随机树在无障碍环境下的扩展速度,并在随机树扩展的过程中加入相关约束条件以确保生成路径的可行性.在找到第一条可达路径后,采用变化的椭圆或椭球采样域限制采样点选取和自适应步长的扩展范围,加快算法收敛到渐进最优的速度.最后,在复杂二维和三维环境下进行传统算法和改进算法的对比实验,仿真分析表明:改进算法可以在很少的迭代次数下找到更优的初始路径,更快地锁定椭圆或椭球采样域,从而给路径优化留出更多时间,算法规划效果更好.

DPCA-POA-RF-Informer在多情景光伏多步预测中的应用

针对光伏发电波动性与不确定性对电力系统稳定产生的影响,对多情景光伏发电功率的多步预测进行研究。首先通过密度峰值算法根据太阳辐射量、温度、湿度等气象数据对天气状况进行精确分类。其次,为了使模型表现出更好的性能,建立了鹈鹕算法优化随机森林(POA-RF)的因素筛选特征变量,模型用鹈鹕算法对随机森林的决策树数目和深度两个参数进行寻优,加强了因素筛选的有效性。最后,基于Informer模型对不同天气状况的光伏功率进行多步预测。实例计算结果验证了所提模型预测精准度的有效性与精准性。

基于Informer算法的网联车辆运动轨迹预测模型

自动驾驶汽车可以根据轨迹预测算法计算周边车辆的运动轨迹,并作出反应以降低行车风险,而传统的轨迹预测模型在长时间序列预测的情况下会产生较大的误差。为解决这一问题,提出了一种以Informer算法为基础的轨迹预测模型,并根据公开数据集NGSIM进行实验分析。首先通过对称指数移动平均法(sEMA)对原始数据进行滤波处理,并在原有的Informer编码器中加入了联合归一化层对不同车辆进行特征提取处理,减少了不同车辆之间的运动误差,通过考虑车辆的本身速度信息与周围环境的车辆运动信息,提高了预测精度,最后经过解码器得到未来时刻的车辆轨迹分布。结果表明,模型对车辆的轨迹预测误差在0.5 m以内;通过对轨迹预测的MAE与MSE结果分析可知,预测时间超过0.3 s以后,Informer模型的轨迹预测效果明显优于其他算法,验证了模型和算法的有效性。

Insomnia Burden among Informal Caregivers of Hospitalized Lung Cancer Patients and Its Influencing Factors

Objective This study aimed to reveal the insomnia burden and relevant influencing factors among informal caregivers(ICs)of hospitalized patients with lung cancer.Methods A cross-sectional study on ICs of hospitalized patients with lung cancer was conducted from December 31,2020 to December 31,2021.ICs’burden was assessed using the Caregiver Reaction Assessment(CRA),Hospital Anxiety and Depression Scale(HADS),and Insomnia Severity Index(ISI).Linear and logistic regression models were used to identify the influencing factors.Results Among 289 ICs of hospitalized patients with lung cancer,83(28.72%),53(18.34%),and 14(4.84%)ICs experienced mild,moderate,and severe insomnia,respectively.The scores concerning self-esteem,lack of family support,financial problems,disturbed schedule,and health problems were 4.32±0.53,2.24±0.79,2.84±1.14,3.63±0.77,and 2.44±0.95,respectively.ICs with higher Activities of Daily Living Scale(ADLS)scores were associated with a lower risk of insomnia,with an odd ratio(OR)and 95%confidence interval(CI)of 0.940(0.898–0.983).Among the ICs,female gender(OR=2.597),alcohol consumption(OR=3.745),underlying medical conditions(OR=11.765),long-term caregiving experience(OR=37.037),and higher monthly expenses(OR=5.714)were associated with a high risk of insomnia.Conclusion Of the hospitalized patients with lung cancer,51.9%experienced insomnia.Patients’ADL,ICs gender,alcohol consumption,underlying medical conditions,caregiving duration,and monthly expenses were influencing factors.Therefore,prompt screening and early intervention for ICs of patients with lung cancer is necessary.

基于改进Informer的云计算资源负载预测

负载预测是云计算资源管理中的重要组成部分,准确预测云资源的使用情况可提高云平台性能及防止资源浪费,然而云计算资源使用的动态性和不确定性使得负载预测较为困难,尽管Informer在时序预测领域取得了较好的效果,但未对时间的因果依赖关系加以限制造成未来信息泄露,也未考虑网络深度的增加导致模型性能下降的问题。为解决上述问题,提出一种基于改进Informer的多步负载预测模型(Informer-DCR)。将编码器中各注意力块之间的正则卷积替换为扩张因果卷积,使深层网络中的高层能够接收更大范围的输入信息来提高模型预测精度,并保证时序预测过程的因果性。在编码器中添加残差连接,使网络中低层的输入信息直接传到后续的高层,解决了深层网络退化问题。实验结果表明,Informer-DCR模型在不同预测步长下的平均绝对误差比Informer、时间卷积网络等主流预测模型降低了8.4%~40.0%,并且在训练过程中表现出比Informer更好的收敛性。

基于策略梯度Informer模型的滚动轴承剩余寿命预测

典型的编码器-解码器——Transformer存在二次时间复杂度、高内存使用及模型结构复杂等固有限制,造成Transformer用于滚动轴承剩余寿命(RUL)预测会表现出较低预测精度和较低计算效率的问题。为此,提出一种新型编解码器——策略梯度Informer(PG-Informer)模型,并将其应用于滚动轴承RUL预测。首先,在PG-In-former的新型编解码器体系结构——Informer中设计了概率稀疏自注意力(PSSA)机制替代Transformer中原有的自注意力机制,以提高非线性逼近能力并减少时间和空间复杂度;然后,PG-Informer采用自注意力蒸馏(SAD)操作减少参数维度和参数量,并提高了对时间序列的预测鲁棒性;此外,PG-Informer的生成式解码器对解码输入进行一步解码输出预测结果,无需动态多步解码,提升了对时间序列的预测速度;最后,构造了策略梯度学习算法来提高对PG-Informer参数的训练速度。PG-Informer的以上优势使所提出的基于PG-Informer模型的滚动轴承RUL预测方法可以获得较高的预测精度、较好的鲁棒性和较高的计算效率。对辛辛那提大学智能维护系统中心的1号滚动轴承的RUL预测实验结果表明,所提出方法预测得到的RUL值为963min,其RUL预测误差仅为6.50%,比基于Transformer的RUL预测方法预测精度更高、预测误差更小、鲁棒性更好;所提出方法所耗费的RUL预测时间仅为132.37s,比基于Transformer的RUL预测方法的预测时间更短。以上实验结果验证了所提出方法的有效性。

一种基于STL-Prophet-Informer模型的太阳电池阵多变量趋势预测方法

为了提高太阳电池阵多变量预测的精度,解决阳电池阵遥测参数存在周期波动与增长性互相耦合的问题,提出一种基于STL-Prophet-Informer模型的太阳电池阵多变量预测算法.该算法首先应用局部加权周期趋势分解算法(seasonal and trend decomposition procedure based on loess,STL)对太阳电池阵的多个参数分解为趋势分量、周期分量和残差分量,然后采用对趋势性数据预测效果较好的Prophet预测趋势分量,Informer模型预测周期分量和残差分量,最后将各分量预测结果相加后得到总的太阳电池阵参数预测值.以某卫星太阳电池阵实际遥测数据做算例分析,提出算法的各项误差评价指标和单一的Informer模型、LSTM模型等相比有明显减小,将该组合预测模型用于太阳电池阵多变量参数预测中,可以提高参数预测精度,提升卫星自主运行性能.

基于Informer神经网络的锂离子电池容量退化轨迹预测

通过对锂离子电池容量退化轨迹的精确预测可以大幅提升电池材料的研究效率。针对Transformer网络在锂电池容量退化轨迹预测这种长时间序列预测任务中存在的问题,本工作采用滑动窗口策略,构建了一种基于Informer网络的锂离子电池容量退化轨迹预测方法。首先,利用滑动窗口对数据集进行划分和再拼接,便于神经网络挖掘数据序列内部的相关性;然后,根据Informer网络的周期性时间特征捕捉能力设计适用于锂电池数据的全局时间戳;最后,使用前10%容量数据通过多步滚动预测方法实现模型输出,缓解预测中的误差累积问题,进而得到完整的预测轨迹。通过选取不同的误差评价指标和训练过程中的时间开销,在美国马里兰大学提供的锂电池数据集上验证了所建立模型的准确性和训练效率,并在美国航空航天局提供的电池数据集上验证了模型的泛用性。本工作模型的预测结果与多层感知机神经网络、循环神经网络及Transformer网络模型对比,退化轨迹与真实轨迹最为拟合,且训练时间开销小,预测结果的平均绝对误差和均方根误差控制在2.57%和3.5%,验证了所提预测方法的有效性。

基于GGInformer模型的金融数据特征提取及价格预测

为了解决金融时序预测任务中出现的特征参数冗余问题,用遗传算法对金融数据进行特征提取,通过三组对比实验进行验证分析。实验结果显示,加入了遗传算法的预测模型比未加入遗传算法的模型在三种数据集上的MSE均有所降低。最终结果证明遗传算法可以有效解决金融产品价格预测过程中的特征冗余问题。为了解决非线性的长序列金融数据预测效果差的问题,通过结合GRU网络和Informer模型构建了GGInformer模型来对金融产品价格进行预测。模型在三种外汇产品数据集上与其他四种预测基准方法进行了对比实验,实验结果与可视化分析表明,所提模型在金融产品交易价格的预测结果上有明显优势,可以提高预测的精度。

基于Informer的电离层电子总量预测

电离层会影响卫星导航和通信信号的传播和反射,而电离层总电子含量(TEC)则是评估电离层异常扰动的关键指标,因此,精准预测电离层TEC具有重要意义。本研究从国际GNSS服务(IGS)电离层分析中心获取全球1999—2011年的电离层TEC的时间序列数据,时间分辨率为2小时。选取5个位置,构建了Informer模型预测未来24小时的TEC数据,为研究电离层异常扰动提供科学依据。在多个模型对比中,Informer模型的R-Squared和RMSE表现最佳,验证了Informer方法的可行性。

基于聚合二次模态分解及Informer的短期负荷预测

针对区域级负荷的非平稳性及长序列预测精度低的问题,该文提出了一种基于聚合二次模态分解及Informer的短期负荷预测方法。首先,运用改进完全集合经验模态分解(improved complete ensemble empirical mode decomposition with adaptive noise,ICEEMDAN)对负荷序列进行初步分解,削弱原始序列的随机性与波动性;其次,根据子序列的样本熵计算结果进行聚合,并通过比较不同的聚合方式选出最优重构方案;然后,利用变分模态分解对高复杂度的合作模态函数进行二次分解;充分考虑到电价、气象等因素对负荷的影响,采用随机森林(random forest,RF)算法进行相关性分析,从而为每个子序列构建不同的高耦合度特征矩阵并输入Informer进行建模,并通过其多层次编码及稀疏多头自注意力机制等方式提高对负荷序列的预测效率;最后采用巴塞罗那区域级负荷数据集进行实例验证,结果显示所提框架有效解决了模态分解过程中的模态混叠以及高频分量问题,并且其长序列预测均方根误差相比其他经典深度学习模型最高降低了65.28%。

Escherichia coli Pathotypes in Children with Acute Diarrhea in an Informal Settlement in Nairobi, Kenya

Diarrhea is among the leading causes of morbidity and mortality in children aged Escherichia coli (DEC) accounts for 30% - 40% of childhood diarrhea cases. To identify the pathotypes involved in diarrheal outbreaks in Kenya, we analyzed archived E. coli isolates from children E. coli confirmation and antimicrobial susceptibility testing were done using the VITEK®2 instrument. Pathotype identification was performed via conventional polymerase chain reaction. Of 175 E. coli isolates, 48 (27%) were DEC pathotypes, with enteroaggregative E. coli (EAEC) predominating (71%, 34/48). Enterohemorrhagic (EHEC) and enteropathogenic E. coli (EPEC) represented 19% and 10% of isolates, respectively. Enteroinvasive and enterotoxigenic pathotypes were not identified. All DEC isolates were susceptible to amikacin, ertapenem, imipenem, meropenem and tigecycline. Conversely, most (>80%) isolates were resistant to ampicillin, ampicillin-sulbactam and sulfamethoxazole-trimethoprim. Half of all EAEC and EPEC strains were resistant to cefazolin while half of EHEC isolates were resistant to ciprofloxacin and moxifloxacin. In total, 18 resistance phenotypes were identified with “ampicillin-cefazolin-ampicillin/ sulbactam-sulfamethoxazole/trimethoprim” predominating (33%, 16/48). The majority (81%) of DEC isolates were multidrug-resistant, with extended-spectrum beta-lactamase production identified in 8% of these isolates. This study highlights the predominance of Enteroaggregative E. coli and multidrug resistance of DEC pathotypes. Studying the epidemiology of diarrheal disease and antimicrobial resistance surveillance, will aid in identifying dominant etiological agents of diarrhea and newly emerging resistant strains in informal settlements.

基于LSTM和Informer变体的渔业水质预测研究

现代化渔业养殖需要更快更准确的水质预测,提升养殖水质预测的精度对提升渔业养殖的效率具有重要意义,为此提出一种结合随机森林、SG滤波器、LSTM和Informer变体的渔业水质预测模型。首先通过随机森林算法填补数据中的缺失值,通过SG滤波器减少噪声干扰。其次将LSTM结合静态和瞬时递归网络作为Informer中的内部结构,将数据送入模型中。最后通过全连接层输出得到水质预测结果,对渔场内监测点的水温、pH值、溶解氧进行预测。结果表明所提方法提高了渔业养殖水质预测的预测精度。

Rice vs.Wheat:How Rice Farming Shaped Informal Finance in China

China’s financial system is characterized by the coexistence of both formal and informal finance,and informal finance appears to be more prevalent in the south of the country compared to the north.In this paper,we use the historical demarcation of rice-farming regions and the percentage area of rice sown to ascertain the regional distribution of rice farming in the south and wheat farming in the north in order to analyze the role of rice farming in the evolution of different types of financial institutions in China.We find that rice cultivation has a positive effect on informal financial development,though its impact on formal finance is insignificant.Unlike wheat farming,rice cultivation requires farmers to collaborate during busy farming seasons and necessitates the construction of irrigation systems.Such collaboration helped enhance an acquaintance society,in which both parties to transactions enforced the rules and contract terms of their own initiative,paving the way for informal finance.This paper not only demonstrates the cultural roots of the regional financial disparities in China but also expands the research on the ways in which farming practices from the past have shaped the development of China’s present financial markets.

基于改进Informed-RRT^(*)算法的舰载机甲板平面路径规划

针对舰载机甲板路径规划问题,在Informed-RRT^(*)(informed rapidly-exploring random tree)的椭圆采样基础上,提出使用正态分布方式采样的IN-RRT^(*)(informed normal-RRT^(*))算法。首先,针对舰载机与运动场景建模,定义舰载机运动约束和避障策略;其次,将正态分布采样策略与椭圆采样相结合,获取优质高效采样点;引入人工势场法,自适应调节随机树的搜索步长值;使用向心Catmull-Rom样条插值法对路径进行平滑优化处理;提出针对动态障碍改进的动态窗口法,实现局部动态避障。最后,运用甲板平面环境实验检验算法性能。结果表明,IN-RRT^(*)算法能显著优化搜索时间和搜索路径质量,可应对动态场景规划出合理可行的平滑路径。