Machine-Learning based methods in short-term load forecasting

Short-term load forecasting is of great significance to the secure and efficient operation of power systems. However, loads can be affected by a variety of external impact factors and thus involve high levels of uncertainties. So it is a challenging task to achieve an accurate load forecast. This paper discusses three commonly-used machine-learning methods used for load forecasting, i.e., the support vector machine method, the random forest regression method, and the long short-term memory neural network method. The features and applications of these methods are analyzed and compared. By integrating the advantages of these methods, a fusion forecasting approach and a data preprocessing technique are proposed for improving the forecasting accuracy. A comparative study based on real load data is performed to verify that the proposed approach is capable of achieving a relatively higher forecasting accuracy.     

6UPS并联机构杆长误差标定及其对动平台运动的影响分析

分析了不同误差对并联机构的影响,尤其是杆长误差对动平台姿态误差的影响.建立了采用并联机构运动学反解模型的矢量法误差求解模型;针对6UPS并联刀具机构的结构特点,使用支链分析方法,对由杆长误差引起的动平台误差变化趋势及变化的敏感性方向进行了分析.使用激光干涉仪完成了误差测量实验,使用多组姿态数据的不同组合进行误差计算,获得了优化的杆长误差数据.完成了杆长误差对动平台误差变化影响的数值分析与矢量合成两种方法的验证.     

基于邻域彩色变化矢量场的图像边缘检测技术研究*

首先进行了边缘检测系统结构设计,建立了图像邻域彩色变化矢量场的数理模型,提出了用图像邻域彩色变化方向锐度描述图像边缘,进而应用模糊聚类自适应检测边缘.实验表明:与基于梯度的边缘检测技术相比,该方法在噪声抑制以及边缘准确定位上均取得了好的效果,是一种应用广泛的优秀边缘检测算法.     

Development of Covalent Chitosan-Polyethylenimine Derivatives as Gene Delivery Vehicle: Synthesis,Characterization, and Evaluation

There is an increasing interest in cationic polymers as important constituents of non-viral gene delivery vectors. In the present study, we developed a versatile synthetic route for the production of covalent polymeric conjugates consisting of water-soluble depolymerized chitosan (dCS; M_W 6–9 kDa) and low molecular weight polyethylenimine (PEI; 2.5 kDa linear, 1.8 kDa branched). dCS-PEI derivatives were evaluated based on their physicochemical properties, including purity, covalent bonding, solubility in aqueous media, ability for DNA condensation, and colloidal stability of the resulting polyplexes. They were complexed with non-integrating DNA vectors coding for reporter genes by simple admixing and assessed in vitro using liver-derived HuH-7 cells for their transfection efficiency and cytotoxicity. Using a rational screening cascade, a lead compound was selected (dCS-Suc-LPEI-14) displaying the best balance of biocompatibility, cytotoxicity, and transfection efficiency. Scale-up and in vivo evaluation in wild-type mice allowed for a direct comparison with a commercially available non-viral delivery vector (in vivo-jetPEI). Hepatic expression of the reporter gene luciferase resulted in liver-specific bioluminescence, upon intrabiliary infusion of the chitosan-based polyplexes, which exceeded the signal of the in vivo jetPEI reference formulation by a factor of 10. We conclude that the novel chitosan-derivative dCS-Suc-LPEI-14 shows promise and potential as an efficient polymeric conjugate for non-viral in vivo gene therapy.     

Adenovirus-Mediated Inducible Expression of a PD-L1 Blocking Antibody in Combination with Macrophage Depletion Improves Survival in a Mouse Model of Peritoneal Carcinomatosis

Immune checkpoint inhibitors (ICIs) have demonstrated remarkable efficacy in a growing number of malignancies. However, overcoming primary or secondary resistances is difficult due to pharmacokinetics issues and side effects associated with high systemic exposure. Local or regional expression of monoclonal antibodies (mAbs) using gene therapy vectors can alleviate this problem. In this work, we describe a high-capacity adenoviral vector (HCA-EFZP-aPDL1) equipped with a mifepristone-inducible system for the controlled expression of an anti-programmed death ligand 1 (PD-L1) blocking antibody. The vector was tested in an immune-competent mouse model of colorectal cancer based on implantation of MC38 cells. A single local administration of HCA-EFZP-aPDL1 in subcutaneous lesions led to a significant reduction in tumor growth with minimal release of the antibody in the circulation. When the vector was tested in a more stringent setting (rapidly progressing peritoneal carcinomatosis), the antitumor effect was marginal even in combination with other immune-stimulatory agents such as polyinosinic-polycytidylic acid (pI:C), blocking mAbs for T cell immunoglobulin, mucin-domain containing-3 (TIM-3) or agonistic mAbs for 4-1BB (CD137). In contrast, macrophage depletion by clodronate liposomes enhanced the efficacy of HCA-EFZP-aPDL1. These results highlight the importance of addressing macrophage-associated immunoregulatory mechanisms to overcome resistance to ICIs in the context of colorectal cancer.     

基于三维特征向量的非侵入式电热负荷细分算法

非侵入式负荷监测分解（NILMD）技术是当前居民用能服务深化提升和电力供需互动的重要数据获取手段,然而当前工程上应用广泛的事件驱动型NILMD技术一直无法准确细化分解电热负荷。针对这一问题,文中提出了一种基于三维特征向量的典型电热负荷细化分解算法。首先,基于有功、无功功率和电流谐波等电气负荷特征采用事件检测方法提取电热事件,在有功功率的基础上,引入运行时长、频繁启停次数等非电气负荷特征共同构建三维特征向量电器模型。然后,采用序贯覆盖法设计典型电热负荷细化分解命题学习规则和细化分解算法。最后,基于实证实验数据进行分解验证,发现4种典型电热负荷的细化分解准确率超过85%。实验结果表明,文中所提典型电热负荷细化分解算法有效地提高了4种典型电热负荷分解的准确率。     

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。     

基于图像识别的铜转炉吹炼造渣期渣含Fe预测模型研究

铜转炉吹炼是火法炼铜的关键工序,其终点判断与炉寿、铜产率和直收率紧密相关,目前现有人工经验、仪器测定和物料平衡法等终点判断方法均存在一定的局限性。理论上铜转炉吹炼造渣期终点与渣含Fe是否达标有关,而不同Fe含量渣样呈现不同的图像特征,鉴于此,基于图形识别的特征向量提取原理,分别采用卷积神经网络(CNN)算法与支持向量机(SVM)算法,构建了铜转炉吹炼造渣期渣含Fe预测模型,为图像识别技术在铜转炉吹炼终点判断中的应用奠定数模基础。两种模型的实例分析表明,卷积神经网络的训练集预测准确率98%,测试集预测准确率约50%;支持向量机模型的训练集预测准确率99%,测试集预测准确率62%。     

Quadratic optimization fine tuning for the Support Vector Machines learning phase

This work presents a comparative analysis of specific, rather than general, mathematical programming implementation techniques of the quadratic optimization problem (QP) based on Support Vector Machines (SVM) learning process. Considering the Karush–Kuhn–Tucker (KKT) optimality conditions, we present a strategy of implementation of the SVM-QP following three classical approaches: (i) active set, also divided in primal and dual spaces, methods, (ii) interior point methods and (iii) linearization strategies. We also present the general extension to treat large-scale applications consisting in a general decomposition of the QP problem into smaller ones, conserving the exact solution approach. In the same manner, we propose a set of heuristics to take into account for a better than a random selection process for the initialization of the decomposition strategy. We compare the performances of the optimization strategies using some well-known benchmark databases.