基于传递熵与KELM的炼油化工过程风险传播路径分析方法

基于炼油化工过程复杂,设备众多,某一设备的监测变量发生扰动可能会传播至其相邻设备引发出一系列故障链。现有方法多是针对某一设备进行监测与诊断,以期降低事故后果,而忽视了对过程风险传播路径的预测以防止事故的发生。因此,提出一种基于传递熵与核极限学习机的炼油化工过程风险传播路径分析方法,该方法针对某一工艺扰动,分析其在风险发展过程中的扰动传播过程,基于传递熵分析法建立炼油化工过程风险传播推绎模型;并提出一种基于KELM的风险传播搜索方法,预测风险传播路径;将该方法应用于分馏塔冲塔过程。研究结果表明:该方法可辨识出未来一段时间内风险的可能传播路径,以便操作人员及时采取预防措施,保证过程安全及产品质量。     

Enduring Extremes? Polar Vortex,Drought, and Climate Change Beliefs

Some extreme weather events may be more likely to affect climate change beliefs than others, in part because schema individuals possess for different events could vary in encouraging such links. Using a representative sample of U.S. adults and geocoded National Weather Service data, we examine how a range of extreme weather event categories relate to climate change beliefs, and the degree to which individuals’ self-reported experiences are shaped by their political views across event types. For tornado, hurricane, and flood events, we find no link with beliefs. For polar vortex and drought events, we find that although self-reported experience is linked with climate beliefs, reporting of these experiences is influenced by political identity and partisan news exposure. These findings underscore a limited role for extreme weather experiences in climate beliefs, and show that events more open to interpretation, such as droughts and polar vortex disturbances, are most likely to be seen through a partisan lens.     

基于有上下界风量约束的矿井风量极值流算法研究

为解决既有方法不能准确求出矿井最大通风量与最小配风量,从而导致矿井通风系统产生能源浪费或难以满足矿井通风需求,出现安全隐患与改造难度大、潜力低等问题,提出1种基于有上下界风量约束的矿井风量极值流算法.该算法首先利用最大流算法求解满足节点流量平衡与分支上下界容量限制条件的可行流问题,若存在可行流,在求得可行流基础上,再利...     

Trends in Raindrop Kinetic Energy with Modeled Climate Warming in the Lake Tahoe Basin

Two means by which climate change may increase surface soil erosion in mountainous terrain are: (1) increasing the frequency of extreme rainfall events and (2) decreasing the duration of snow cover on bare soil. We used output from four general circulation models (GCMs) and two greenhouse gas trajectories to produce a suite of hydrologic variables at a daily time‐step for historic and projected 21st Century conditions. We statistically disaggregated the daily rainfall to hourly, using hourly rainfall from a network of nine weather stations in the Tahoe Basin, and filtered out rain falling on a snowpack. We applied published equations to convert hourly intensity to raindrop kinetic energy (KE) for each day and grid cell in the Basin, averaged across grid cells, and created time series of total annual and maximum annual hourly kinetic energy (TKE and MKE) on snow‐free ground. Using the Generalized Extreme Value distribution, we calculated the significance of long‐term trends in KE on snow‐free ground, and estimated energy levels for return periods of 2, 20, and 100 years. We then detrended the snowpack data and compared the resulting trends in KE with the trends resulting from changes in both rainfall energy and snowpack under two GCMs. Principal findings include (1) upward trends in MKE, (2) stronger upward trends in TKE; and (3) an effect of increasing rainfall intensities on KE in some cases, and a strong effect of reduced snowpack in all cases examined.     

The effect of summer drought on the predictability of local extinctions in a butterfly metapopulation

The ecological impacts of extreme climatic events on population dynamics and community composition are profound and predominantly negative. Using extensive data of an ecological model system, we tested whether predictions from ecological models remain robust when environmental conditions are outside the bounds of observation. We observed a 10-fold demographic decline of the Glanville fritillary butterfly (Melitaea cinxia) metapopulation on the Åland islands, Finland in the summer of 2018 and used climatic and satellite data to demonstrate that this year was an anomaly with low climatic water balance values and low vegetation productivity indices across Åland. Population growth rates were strongly associated with spatiotemporal variation in climatic water balance. Covariates shown previously to affect the extinction probability of local populations in this metapopulation were less informative when populations were exposed to severe drought during the summer months. Our results highlight the unpredictable responses of natural populations to extreme climatic events.     

SVD在串联故障电弧检测及选相中的应用*

煤矿井下发生串联故障电弧易引发火灾等安全事故,为了预防电气火灾、指导线路维修,利用三相电动机及变频器负载开展不同线路、不同电流条件下的串联故障电弧实验,研究三相串联故障电弧的检测及选相方法。首先,对单相电流进行一阶差分处理后,建立改进的吸引子轨迹矩阵作为故障特征矩阵;其次,对故障特征矩阵进行奇异值分解,采用特征矩阵的奇异值构建串联故障电弧检测及选相的特征向量;最后,利用极限学习机建立故障电弧检测及选相模型,并测试检测及选相准确率。研究结果表明:提出的SVD方法可以利用单相电流实现三相电动机及变频器负载回路中的串联故障电弧检测及选相。     

基于标准化降水蒸散指数的华南干旱趋势研究

最近几年华南地区干旱频发,为探讨该地区的干旱趋势,用标准化降水蒸散指数(SPEI)和1961-2010年华南地区具有代表性的50个站点的月降水及月平均气温资料,分析了该地区近50年来的干旱趋势、干旱空间分布、极端干旱事件发生频次和干旱持续时间。结果表明,华南地区普遍存在干旱事实,最近10年是干旱最严重的10年,Mann-Kendall检验表明该地区平均SPEI指数从1998年开始突变;干旱化最严重的区域是海南岛、广西南部和西部地区,广东的干旱化趋势最轻。20世纪70年代干旱和极端干旱事件较少,其后明显增多,干旱持续时间也有所延长。由于该地区降水呈现弱增加趋势而温度升高显著,因此推测温度升高导致蒸散增加可能是华南地区干旱化的主要原因。另外,降水频次的减少和集中也是导致近来极端干旱事件增多的原因之一。SPEI指数较好地体现了气候变暖导致的干旱化趋势。     

长江三峡库区极端大雾天气的气候变化特征

为了进一步认识长江三峡库区大雾天气特征,采用线性趋势估计和Molet小波分析方法,研究了库区持续12 h以上和连续3 d以上极端大雾天气气候变化特征,探讨了库区蓄水后大雾天气气候变化的原因。结果表明：三峡库区年平均雾日数呈弱的下降趋势,存在准8、18、32 a的年代际周期振荡。持续12 h以上和连续3 d以上大雾天气有明显增加趋势,分别存在准10、17、32 a和准12、32 a的年代际周期振荡。在蓄水后,库区西段年平均雾日数明显减少、东段略有增加,持续12 h以上大雾年平均日数变化不大,连续3 d以上大雾年平均日数明显减少。库区年平均雾日数的总体减少在很大程度上是受全球气候变暖以及城市化共同影响的结果,没有证据说明三峡库区蓄水对大雾天气有明显影响。     

近50年宁夏极端气温事件的变化研究

随着全球变暖,极端气候事件频繁发生,由此造成气象灾害的数量日益增加。深入研究极端气候的变化特征,能够为预测和预防极端事件灾害提供参考依据。采用线性倾向估计法、反距离加权法和R/S分析法,选取10个极端气温指标研究了宁夏近50年来极端气温事件的时空变化特征,并在此基础上尝试预测了未来该地区极端气温变化的情形。结果发现:全天极端高温天数、白天极端高温天数、夜间极端高温天数、生物生长季和夏季天数分别以0.76、0.48、0.67、0.35和0.29 d/a的趋势明显增加,而全天极端低温天数、白天极端低温天数、夜间极端低温天数和最大连续霜冻天数分别以-0.40、-0.25、-0.66和-0.30 d/a趋势显著减少,极端气温年较差也呈下降趋势(-0.02℃/a),且空间差异明显;除极端气温年较差外,其它各极端气温指标与年平均气温均有很好的相关性;年极端冷指标和极端气温年较差在未来将继续下降,极端暖指标在未来将继续上升;宁夏气象灾害所造成的影响和损失将进一步增大。     

基于可解释机器学习的青藏高原草地物候变化多因素影响分析

气候变化背景下,青藏高原植被物候发生显著改变.然而,影响物候的水热因素众多,目前较少有研究关注多因素对青藏高原物候的影响效应,导致对青藏高原物候变化机制认识不足.为此,研究通过遥感数据解译,在对2002~2021年青藏高原草地物候时空变化特征分析的基础上,聚焦降水、气温、海拔和土壤等多方面,利用可解释机器学习方法（SHAP）揭示物候变化的主导因素,并量化分析多因素对物候的交互影响.结果表明:①青藏高原分别有56.32 %、67.65 %和65.50 %的草地表现出生长季开始时间（SOS）提前、生长季结束时间（EOS）延迟和生长季长度（LOS）延长趋势;②青藏高原草地SOS和LOS主要受水分条件影响,3月0~10 cm土壤水分对SOS提前和LOS延长起促进作用的范围分别在10~25 kg·m^-2和15~25 kg·m^-2之间,峰值分别在20 kg·m^-2和18 kg·m^-2左右;EOS则主要受温度影响,9月和10月温度越高对EOS延迟促进作用越强,并分别在高于8 ℃和-0.5 ℃时达到峰值;③水热等因素对物候的影响存在非线性交互效应,3月0~10 cm土壤水分达到20 kg·m^-2后,更有利于低降水和低海拔地区SOS提前;10月温度高于0 ℃后较好的水分条件更有利于EOS延迟;3月0~10 cm土壤水分在12~22 kg·m^-2之间时,高降水地区LOS更长.研究表明,可解释机器学习方法可为物候变化的多因素影响定量分析提供一种新的方法.