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。     

成都市臭氧重污染成因与来源解析

采用来源解析的方法对2020年成都市发生的一次较为严重的臭氧污染事件进行了研究.结果表明,此次污染过程呈现从清洁-污染-清洁的变化趋势,污染持续时间长达9d,最大臭氧小时浓度达到258.8μg/m3.气象因素在成都臭氧污染中的影响不可忽略,其中温度与臭氧浓度呈现显著正相关关系,东北风主要出现在污染前和污染后,可能起到稀...     

SPE-UPLC-MS/MS法测定地下水中多种农药残留

采用固相萃取超高效液相色谱串联质谱法(SPE-UPLC-MS/MS)同时测定地下水中涕灭威、克百威、2,4滴和五氯酚等4种农药残留,通过优化试验条件,使方法在10.0μg/L～200μg/L范围内线性良好,检出限为0.003μg/L～0.006μg/L.标准溶液低、中2个质量浓度水平的加标回收率为84.1％ ～98.8...     

西江溶解有机碳的输送对典型洪水过程的响应

对2005年6月西江干流出现特大洪水期间及其前后的溶解有机碳(DOC)进行连续采样分析.结果表明,DOC浓度对流量变化的响应不敏感(n=43,R2=0.23,p＞0.05),尽管流量有6～7倍的变化,但DOC浓度的变化不到30%.洪水前夕、洪水期间及洪水过后DOC平均浓度逐渐降低,且洪水过后西江下游各个采样点的DOC浓度明显低于洪水前夕.DOC的输出量主要受地表径流深度的影响,二者之间存在着显著的线性相关关系.DOC荷载通量与流量密切相关,通过建立二者之间的线性回归方程可以计算出DOC的输出量,洪水期间DOC的输出量为4.5×1010g.     

Some properties of a sequencing batch reactor for treatment of wastewater containing thiocyanate compounds

Thiocyanate (SCN) compounds in photo-processing wastewater (PPWW) could be treated by an SBR system without any release of thiocyanate to the atmosphere during the aeration step. An SCN loading greater than 84 g m(-3)d(-1) showed negative effects on the growth of bio-sludge and removal efficiencies of the system. The acclimatization period of the system was increased with an increase in SCN concentration or loading. The COD, BOD(5), TKN, and SCN removal efficiencies were 96.0 +/- 1.6%, 72 +/- 2%, 49 +/- 5%, and 82 +/- 3%, respectively, under an SCN loading of up to 84 g m(-3)d(-1). The removal efficiency of the system was repressed by SCN due to the repressed growth rate of nitrification bacteria. However, the removal efficiency could be increased with an increase in HRT or a decrease in SCN loading. Also, increases in HRT or decreases in SCN loading led to increased sludge age or solid retention time (SRT) and decreased the sludge volume index (SVI) value. The SRT and SVI of the system with synthetic wastewater containing 840 mg l(-1) SCN under an HRT of 3 days (SCN loading of 280 g m(-3)d(-1)) were 3.9 +/- 0.7 days and 65 +/- 4 ml g(-1), respectively, while they were 11.2+/-0.8 days and 55 +/- 6 ml g(-1), respectively under an HRT of 10 days (SCN loading of 84 g m(-3)d(-1)).     

Response of Fish and Macroinvertebrate Bioassessment Indices to Water Chemistry in a Mined Appalachian Watershed

Multimetric indices based on fish and benthic macroinvertebrate assemblages are commonly used to assess the biological integrity of aquatic ecosystems. However, their response to specific stressors is rarely known. We quantified the response of a fish-based index (Mid-Atlantic Highlands Index of Biotic Integrity, MAH-IBI) and a benthic invertebrate-based index (West Virginia Stream Condition Index, WV-SCI) to acid mine drainage (AMD)-related stressors in 46 stream sites within the Cheat River watershed, West Virginia. We also identified specific stressor concentrations at which biological impairment was always or never observed. Water chemistry was extremely variable among tributaries of the Cheat River, and the WV-SCI was highly responsive across a range of AMD stressor levels. Furthermore, impairment to macroinvertebrate communities was observed at relatively low stressor concentrations, especially when compared to state water quality standards. In contrast to the WV-SCI, we found that the MAH-IBI was significantly less responsive to local water quality conditions. Low fish diversity was observed in several streams that possessed relatively good water quality. This pattern was especially pronounced in highly degraded subwatersheds, suggesting that regional conditions may have a strong influence on fish assemblages in this system. Our results indicate that biomonitoring programs in mined watersheds should include both benthic invertebrates, which are consistent indicators of local conditions, and fishes, which may be indicators of regional conditions. In addition, remediation programs must address the full suite of chemical constituents in AMD and focus on improving linkages among streams within drainage networks to ensure recovery of invertebrate and fish assemblages. Future research should identify the precise chemical conditions necessary to maintain biological integrity in mined Appalachian watersheds.     

Assessing the economic impact of invasive species: the case of yellow starthistle (Centaurea solsitialis L.) in the rangelands of Idaho, USA

Yellow starthistle (Centaurea solsitialis L.) is an invasive weed that creates problems for the management of Idaho's rangelands. A bioeconomic approach combined with an input-output economic model is used to estimate direct and secondary economic costs of the weed in relation to its interference with agricultural and non-agricultural benefits that rangelands provide. Direct economic costs of the infestations were estimated to be of 8.2 million '05 dollars per year, and secondary costs of 4.5 million '05 dollars per year, for a total of 12.7 million '05 dollars; agricultural related economic impacts accounted for 79% of this total cost, and non-agricultural for 21%.     

Is Municipal Solid Waste Recycling Economically Efficient?

It has traditionally been argued that recycling municipal solid waste (MSW) is usually not economically viable and that only when externalities, long-term dynamic considerations, and/or the entire product life cycle are taken into account, recycling becomes worthwhile from a social point of view. This article explores the results of a wide study conducted in Israel in the years 2000–2004. Our results reveal that recycling is optimal more often than usually claimed, even when externality considerations are ignored. The study is unique in the tools it uses to explore the efficiency of recycling: a computer-based simulation applied to an extensive database. We developed a simulation for assessing the costs of handling and treating MSW under different waste-management systems and used this simulation to explore possible cost reductions obtained by designating some of the waste (otherwise sent to landfill) to recycling. We ran the simulation on data from 79 municipalities in Israel that produce over 60% of MSW in Israel. For each municipality, we were able to arrive at an optimal method of waste management and compare the costs associated with 100% landfilling to the costs born by the municipality when some of the waste is recycled. Our results indicate that for 51% of the municipalities, it would be efficient to adopt recycling, even without accounting for externality costs. We found that by adopting recycling, municipalities would be able to reduce direct costs by an average of 11%. Through interviews conducted with representatives of municipalities, we were also able to identify obstacles to the utilization of recycling, answering in part the question of why actual recycling levels in Israel are lower than our model predicts they should be.     

Application of Zoning and “Limits of Acceptable Change” to Manage Snorkelling Tourism

Zoning and applying Limits of Acceptable Change (LAC) are two promising strategies for managing tourism in Marine Protected Areas (MPAs). Typically, these management strategies require the collection and integration of ecological and socioeconomic data. This problem is illustrated by a case study of Koh Chang National Marine Park, Thailand. Biophysical surveys assessed coral communities in the MPA to derive indices of reef diversity and vulnerability. Social surveys assessed visitor perceptions and satisfaction with conditions encountered on snorkelling tours. Notably, increased coral mortality caused a significant decrease in visitor satisfaction. The two studies were integrated to prescribe zoning and “Limits of Acceptable Change” (LAC). As a biophysical indicator, the data suggest a LAC value of 0.35 for the coral mortality index. As a social indicator, the data suggest that a significant fraction of visitors would find a LAC value of under 30 snorkellers per site as acceptable. The draft zoning plan prescribed four different types of zones: (I) a Conservation Zone with no access apart from monitoring or research; (II) Tourism Zones with high tourism intensities at less vulnerable reefs; (III) Ecotourism zones with a social LAC standard of <30 snorkellers per site, and (IV) General Use Zones to meet local artisanal fishery needs. This study illustrates how ecological and socioeconomic field studies in MPAs can be integrated to craft zoning plans addressing multiple objectives.