基于模糊聚类的马氏链模型在交通事故预测中的应用

运用模糊聚类分析方法,借助于MATLAB软件,对中国31个地区交通事故的危害程度进行动态分类和综合评价。根据中国1999—2005年31个地区的交通事故4个指标和5个社会经济因素数据,把31个地区分为3类,即轻灾区、较重灾区、重灾区;其中浙江、福建与宁夏的交通事故危害严重,上海、北京与天津的交通状况正在好转,这与客观实际符合。根据模糊聚类的结果,求得交通事故状态转移概率矩阵,利用马氏链模型可预测未来31个地区的交通事故发展趋势。     

灾变预测在企业安全事故预测中的应用

简述了灰色理论中灾变预测的特点及其建模过程,并将其应用于企业安全事故预测之中.经检验,模型的精度等级为好,模型的预测结果可靠,对于企业安全事故的预测有一定帮助.     

SPATIAL VARIATION OF PRECIPITATION ON THE TEXAS HIGH PLAINS1

ABSTRACT: Space autocorrelation techniques have been used to reveal the nature and spatial distribution of precipitation in the Texas High Plains. Correlation in precipitation amounts varies with both distance and direction, dropping off rapidly with distance, particularly during the warm season. The analyses can be used to estimate storm characteristics in conjunction with a wide variety of problems dealing with irrigation, crop yields, drainage and water supply, and evaluation of artificial weather modification efforts.     

EFFECTS OF THE URBAN ENVIRONMENT ON HEAVY RAINFALL DISTRIBUTION1

ABSTRACT: A network of 225 recording raingages was operated over an area of 5200 km² in the St. Louis region during 1971-1975, in conjunction with an extensive investigation of urban effects on precipitation. Study of urban-induced effects on the frequency of heavy rainstorms has revealed a pronounced increase in the occurrence of storms producing 25 mm (1 inch) or more of rain. The increase is greatest in an area that is frequently in the path of storms passing across two urban-industrial regions. Analyses of raincells (rain intensity centers) within heavy convective storms shows a pronounced increase in water yield from cells exposed to potential urban effects, compared with those exposed only to the surrounding rural environment. Naturally-occurring heavy cells tend to undergo the greatest enhancement from urban exposure. Other analyses indicate an above-average frequency of excessive rain rates for periods of five minutes to two hours downwind of the urban-industrial complex. It is concluded that urban-induced intensification of short-duration rainstorms is sufficient to merit inclusion in the design and operation of urban-area hydrologic systems that control the flow of surplus storm water.     

China's Present Situation of Coal Consumption and Future Coal Demand Forecast

This article analyzes China's coal consumption changes since 1991 and proportion change of coal consumption to total energy consumption. It is argued that power, iron and steel, construction material, and chemical industries are the four major coal consumption industries, which account for 85% of total coal consumption in 2005. Considering energy consumption composition characteristics of these four industries, major coal demand determinants, potentials of future energy efficiency improvement, and structural changes, etc., this article makes a forecast of 2010s and 2020s domestic coal demand in these four industries. In addition, considering such relevant factors as our country's future economic growth rate and energy saving target, it forecasts future energy demands, using per unit GDP energy consumption method and energy elasticity coefficient method as well. Then it uses other institution's results about future primary energy demand, excluding primary coal demand, for reference, and forecasts coal demands in 2010 and 2020 indirectly. After results comparison between these two methods, it is believed that coal demands in 2010 might be 2620-2850 million tons and in 2020 might be 3090-3490 million tons, in which, coal used in power generation is still the driven force of coal demand growth.     

Long‐Term Dynamics of a Fragmented Rainforest Mammal Assemblage

Abstract: Habitat fragmentation is a severe threat to tropical biotas, but its long‐term effects are poorly understood. We evaluated longer‐term changes in the abundance of larger (>1 kg) mammals in fragmented and intact rainforest and in riparian “corridors” in tropical Queensland, with data from 190 spotlighting surveys conducted in 1986–1987 and 2006–2007. In 1986–1987 when most fragments were already 20–50 years old, mammal assemblages differed markedly between fragmented and intact forest. Most vulnerable were lemuroid ringtail possums (Hemibelideus lemuroides), followed by Lumholtz's tree‐kangaroos (Dendrolagus lumholtzi) and Herbert River ringtail possums (Pseudocheirus herbertensis). Further changes were evident 20 years later. Mammal species richness fell significantly in fragments, and the abundances of 4 species, coppery brushtail possums (Trichosurus vulpecula johnstoni), green ringtail possums (Pseudochirops archeri), red‐legged pademelons (Thylogale stigmatica), and tree‐kangaroos, declined significantly. The most surprising finding was that the lemuroid ringtail, a strict rainforest specialist, apparently recolonized one fragment, despite a 99.98% decrease in abundance in fragments and corridors. A combination of factors, including long‐term fragmentation effects, shifts in the surrounding matrix vegetation, and recurring cyclone disturbances, appear to underlie these dynamic changes in mammal assemblages.     

Development and Operational Testing of a Super‐Ensemble Artificial Intelligence Flood‐Forecast Model for a Pacific Northwest River

Coastal catchments in British Columbia, Canada, experience a complex mixture of rainfall‐ and snowmelt‐driven contributions to flood events. Few operational flood‐forecast models are available in the region. Here, we integrated a number of proven technologies in a novel way to produce a super‐ensemble forecast system for the Englishman River, a flood‐prone stream on Vancouver Island. This three‐day‐ahead modeling system utilizes up to 42 numerical weather prediction model outputs from the North American Ensemble Forecast System, combined with six artificial neural network‐based streamflow models representing various slightly different system conceptualizations, all of which were trained exclusively on historical high‐flow data. As such, the system combines relatively low model development times and costs with the generation of fully probabilistic forecasts reflecting uncertainty in the simulation of both atmospheric and terrestrial hydrologic dynamics. Results from operational testing by British Columbia's flood forecasting agency during the 2013‐2014 storm season suggest that the prediction system is operationally useful and robust.     

基于BERT-BILSTM-CRF模型的电力行业事故文本智能分析*

为解决电力行业事故报告文本较长、语义复杂,难以进行有效文本识别问题,提出1种以BERT作为底层的预训练模型,并设计1种双重注意力机制编码器,结合BILSTM-CRF深度挖掘事故文本语义特征,从而实现文本智能分析。首先构建电力词典,通过对BERT预训练,进行BIO标注,然后引入BILSTM-CRF模型实现对文本标签智能分类,最后将该模型与现行其他4种深度学习模型进行对比。研究结果表明:该模型智能识别精确率、召回率及F1值(查准率)均达到约97%,较其他4种模型中效果最好的模型分别提高0.02,0.03,0.02。研究结果可为电力行业事故报告文本分析提供1种新思路。     

基于缓冲区重采样的LSTM林火预测模型*

为提高林火风险预测精度,挖掘地图上隐含的空间信息、时间序列上隐含的长期趋势和循环波动,提出1种基于缓冲区重采样的长短期记忆（LSTM）林火预测模型,选取15个与林火相关的影响因素,以方差膨胀因子为评价指标对其进行多重共线性检验,方差膨胀因子大于10的因素具有共线性,并采用信息增益率验证筛选结果的合理性。考虑到火灾的空间聚集特性,采用缓冲区分析与过采样相结合方法减少样本不均衡现象的影响,最终得到176 732条样本。对12个影响因素和研究时间段的火点建立LSTM预测模型,对森林火灾发生风险进行预测。研究结果表明:基于缓冲区重采样的LSTM林火预测模型有效考虑时空上隐含的信息,预测模型准确率为87.06%,特异性为97.99%,敏感度为76.12%,阳性预测率为97.43%,阴性预测率为80.41%,ROC曲线与AUC值均优于随机森林（RF）和支持向量机（SVM）这2种基准算法。维尔克松秩和检验发现,本文提出的模型与基准算法结果具有显著性差异。研究结果可为提高林火风险预测精度提供参考。     

2021年中国北方首次沙尘天气的多源遥感分析

利用多源遥感数据和NCEP再分析资料,从沙尘源区的地理环境和气候特征出发,对2021年中国北方首次沙尘天气事件每日的大气环流形势以及沙尘的水平,垂直分布特征进行深入的研究分析.结果表明:源区内异常增温,降水稀少的气候背景下致使大面积裸露松散的土壤含水量较低,为大范围,高强度沙尘天气的形成提供了物质基础;频繁活动的冷空气,是沙尘天气爆发的动力因子.沙尘在强风中沿东南方向向下游地区输送和扩散,中国西北,华北,黄淮,江淮,江汉地区和江南北部等地先后受到沙尘天气的影响,空气质量迅速恶化,首要污染物为PM₁₀.此外,沙尘气溶胶东移也波及朝鲜半岛,日本等下游地区.在输送过程中,内陆地区沙尘主要分布在1~6km,而下游地区的沙尘则集中分布在2km高度附近,粒径较大的沙尘出现在近地表的频率较高,较小的颗粒主要分布在对流层中下层.