基于多模型融合的CNN-LSTM-XGBoost短期电力负荷预测方法

短期电力负荷的精准预测可以有效指导机组组合调度、经济调度与电力市场运营。针对输入数据特征量受限时负荷预测的低精度问题,提出一种基于多模型融合的CNN-LSTM-XGBoost短期电力负荷预测方法。通过建立融合局部特征预提取模块的LSTM（long short term memory）网络结构,并将其与XGBoost（eXtreme boosting system）预测模型并行结合,之后结合MAPE-RW（mean absolute percentage error-reciprocal weight）算法进行模型融合初始权重设置,对最佳权重进行搜索,构建最佳融合模型。通过运用电力负荷数据对所提方法进行预测实验,结果表明CNN-LSTM- XGBoost模型的MAPE（mean absolute percentage error）与RMSE（root mean square error）分别为0.377%与148.419 MW,相比于单一网络模型与融合模型结构实现了误差指标的显著降低,验证了基于多模型融合的CNN-LSTM-XGBoost短期电力负荷预测方法具有较快的模型训练速度、较高的预测准确度与较低的预测误差。     

基于长短期记忆网络的电网故障区域定位与故障传播路径推理

为了在发生故障后维持电力系统的安全稳定,有必要实现对故障区域的快速定位并确定故障冲击的传播路径,提出基于长短期记忆网络(LSTM)的故障区域定位和故障传播路径推理方法.首先,利用LSTM建立2个故障诊断模型分别实现在线检测故障时刻和确定故障区域;然后,通过计算故障点附近线路的端口供给能量确定故障冲击的传播路径;最后,以8机36节点电网为例进行算例验证,结果表明所提模型可在发生故障后短时间内检测到故障,给出故障区域和故障冲击传播的路径,且对噪声有较强的鲁棒性.     

Machine Learning Reduced Gene/Non-Coding RNA Features That Classify Schizophrenia Patients Accurately and Highlight Insightful Gene Clusters

RNA-seq has been a powerful method to detect the differentially expressed genes/long non-coding RNAs (lncRNAs) in schizophrenia (SCZ) patients; however, due to overfitting problems differentially expressed targets (DETs) cannot be used properly as biomarkers. This study used machine learning to reduce gene/non-coding RNA features. Dorsolateral prefrontal cortex (dlpfc) RNA-seq data from 254 individuals was obtained from the CommonMind consortium. The average predictive accuracy for SCZ patients was 67% based on coding genes, and 96% based on long non-coding RNAs (lncRNAs). Machine learning is a powerful algorithm to reduce functional biomarkers in SCZ patients. The lncRNAs capture the characteristics of SCZ tissue more accurately than mRNA as the former regulate every level of gene expression, not limited to mRNA levels.     

The Importance of Free Fatty Chain Length on the Lipid Organization in the Long Periodicity Phase

The skin’s barrier ability is an essential function for terrestrial survival, which is controlled by intercellular lipids within the stratum corneum (SC) layer. In this barrier, free fatty acids (FFAs) are an important lipid class. As seen in inflammatory skin diseases, when the lipid chain length is reduced, a reduction in the barrier’s performance is observed. In this study, we have investigated the contributing effects of various FFA chain lengths on the lamellar phase, lateral packing. The repeat distance of the lamellar phase increased with FFA chain length (C20–C28), while shorter FFAs (C16 to C18) had the opposite behaviour. While the lateral packing was affected, the orthorhombic to hexagonal to fluid phase transitions were not affected by the FFA chain length. Porcine SC lipid composition mimicking model was then used to investigate the proportional effect of shorter FFA C16, up to 50% content of the total FFA mixture. At this level, no difference in the overall lamellar phases and lateral packing was observed, while a significant increase in the water permeability was detected. Our results demonstrate a FFA C16 threshold that must be exceeded before the structure and barrier function of the long periodicity phase (LPP) is affected. These results are important to understand the lipid behaviour in this unique LPP structure as well as for the understanding, treatment, and development of inflammatory skin conditions.     

Telomere Length in Metaphase Chromosomes of Human Triploid Zygotes

The human lifespan is strongly influenced by telomere length (TL) which is defined in a zygote—when two highly specialised haploid cells form a new diploid organism. Although TL is a variable parameter, it fluctuates in a limited range. We aimed to establish the determining factors of TL in chromosomes of maternal and paternal origin in human triploid zygotes. Using Q-FISH, we examined TL in the metaphase chromosomes of 28 human triploid zygotes obtained from 22 couples. The chromosomes’ parental origin was identified immunocytochemically through weak DNA methylation and strong hydroxymethylation in the sperm-derived (paternal) chromosomes versus strong DNA methylation and weak hydroxymethylation in the oocyte-derived (maternal) ones. In 24 zygotes, one maternal and two paternal chromosome sets were identified, while the four remaining zygotes contained one paternal and two maternal sets. For each zygote, we compared mean relative TLs between parental chromosomes, identifying a significant difference in favour of the paternal chromosomes, which attests to a certain “imprinting” of these regions. Mean relative TLs in paternal or maternal chromosomes did not correlate with the respective parent’s age. Similarly, no correlation was observed between the mean relative TL and sperm quality parameters: concentration, progressive motility and normal morphology. Based on the comparison of TLs in chromosomes inherited from a single individual’s gametes with those in chromosomes inherited from different individuals’ gametes, we compared intraindividual (intercellular) and interindividual variability, obtaining significance in favour of the latter and thus validating the role of heredity in determining TL in zygotes. A comparison of the interchromatid TL differences across the chromosomes from sets of different parental origin with those from PHA-stimulated lymphocytes showed an absence of a significant difference between the maternal and paternal sets but a significant excess over the lymphocytes. Therefore, interchromatid TL differences are more pronounced in zygotes than in lymphocytes. To summarise, TL in human zygotes is determined both by heredity and parental origin; the input of other factors is possible within the individual’s reaction norm.     

电网故障处置预案文本中的命名实体识别研究

电网故障处置预案是电网故障处置的重要参考,对电网故障处置预案文本中各类电力设备、名称编号等细粒度的关键实体信息进行抽取,是实现计算机学习理解预案内容并进一步支撑故障处置智能化的重要基础。文中提出一种基于深度学习的电网故障处置预案文本命名实体识别技术,首先采用字向量表征预案文本,然后将注意力机制以及双向长短期记忆网络相结合,有所侧重地提取实体词深层字符特征,最后采用条件随机场求解最优序列化的标注。算例表明:文中所提预案文本命名实体识别模型不依赖人工特征,能够自动高效地提取文本特征,准确识别预案文本中细粒度的实体词,满足预案文本中关键实体信息精确定位和识别的要求。     

高效全自动清淤系统在酸性水库的应用

针对 pH ≤ 3 的酸性水库周边地形复杂无法进入大型机械设备、淤泥沉积深夹带有大量织物、库底覆有土工膜易破损等特点,将前置浮船可升降的水面泥浆输送系统与地面泥浆输送系统串联可实现高浓度远距离管道输送,通过全自动监测,采用智能变频控制,保证系统在泥浆浓度可变、液面高度可变的大范围区间内连续且稳定运行。系统在调试运行期间,抽送高浓度泥浆为 19220.88m3,极大提高了酸性水库的清淤效率,降低了清淤成本。     

马钢直弧型板坯连铸机的设计

介绍了马钢 2 5 0 mm× 1 40 0 mm直弧型坯连铸机设计指导思想、连铸机主要技术性能指标、工艺流程、设备组成及采取的技术措施 :无氧化浇注 ,连续弯曲矫直 ,二冷区气雾冷却及自动控制 ,在线短定尺二次切割 ,对称式双向出坯 ,防止测量基准移位和台阶式对弧样板等。     

长输管道焊接缺陷产生原因及预防

通过对长输管道常见焊接缺陷产生类型及原因的分析,总结出减少和避免缺陷产生的有效办法,在施工过程中严格监督落实,可以在很大程度上提高长输管道的焊接一次合格率.