Automated construction of Wi-Fi-based indoor logical location predictor using crowd-sourced photos with Wi-Fi signals

Owing to the recent proliferation of smartphones and the SNS, a large number of images taken by smartphones at various places have been uploaded to SNSs. In addition, smartphones are equipped with various sensors such as Wi-Fi modules that enable us to generate an image associated with the sensory information that represents the context in which the image was captured. This study demonstrates the benefits of images associated with Wi-Fi signals in the automated construction of a Wi-Fi-based indoor logical location classifier that predicts a semantic location label of a user’s position for shopping complexes. In this study, a logical location class refers to the store class label in a shopping complex, such as Starbucks and H&M. Given a collection of images associated with Wi-Fi signals taken at a shopping complex and the complex’s floor plan, the proposed method first estimates the store label at which an image was taken by analyzing the image and crawled online images of branch stores. Then, the 2D coordinates of the images taken at branch stores on the floor coordinate system can be estimated using the floor plan. Subsequently, by using the Wi-Fi signals of the branch store images and their estimated 2D coordinates, we construct a transformation function that maps Wi-Fi signals onto the 2D coordinates, and we adopt this function to predict an indoor location class of an observed Wi-Fi scan from a smartphone possessed by an end user. The proposed transformation function comprises an ensemble of sub-functions designed based on CVAEs. Finally, we demonstrate the effectiveness of the proposed method for three actual shopping complexes.     

融合CNN-BiLSTM-Attention的集成学习价格预测

价格预测对于大宗农产品市场的稳定具有重要意义,但是大宗农产品价格与多种因素有着复杂的相关关系.针对当前价格预测中对数据完整性依赖性强与单一模型难以全面利用多种数据特征等问题,提出了一种将基于注意力机制的卷积双向长短期记忆神经网络(CNN-BiLSTM-Attention)、支持向量机回归(SVR)与LightGBM组合的增强式集成学习方法,并分别在包含历史交易、天气、汇率、油价等多种特征数据的数据集上进行了实验.实验以小麦和棉花价格预测为目标任务,使用互信息法进行特征选择,选择误差较低的CNN-BiLSTM-Attention模型作为基模型,与机器学习模型通过线性回归进行增强式集成学习.实验结果表明该集成学习方法在小麦及棉花数据集上预测结果的均方根误差(RMSE)值分别为12.812, 74.365,较之3个基模型分别降低11.00%, 0.94%、4.44%,1.99%与13.03%, 4.39%,能够有效降低价格预测的误差.     

联合聚类方法和深度学习的混凝土坝变形预测

混凝土大坝变形预测对其安全运行具有重要意义,针对传统分析方法难以捕捉长期序列时序特征从而导致预测精度较低的问题,本文采用麻雀优化算法（SSA）和K调和均值算法（KHM）相结合对监测值进行聚类以捕捉序列时序特征,然后采用自适应噪声完备集合经验模态分解（CEEMDAN）等方法对聚类结果进行降噪处理,最后采用长短期记忆（LSTM）模型对序列进行预测。分析结果表明,本文所提出的聚类方法具有较好的长序列特征识别能力,结合基于CEEMDAN分解方法去除序列中存在的冗余信息,从而使LSTM模型能够更好地捕捉变形值的时序特性,进而提高预测精度。所提模型具有较好的精度和适应性,可为大坝变形预测提供一种有效方法。     

水风光互补适应性调度的集合卡尔曼滤波方法北大核心CSCD

水风光互补是实现能源转型、助力碳中和的重要途径。传统的水风光互补中长期调度模型,忽略了气候变化和人类活动引起的水风光资源变异,得到的静态调度规则难以适应变化环境。为此,本研究采用集合卡尔曼滤波技术开展适应性调度研究。首先进行确定性优化调度,得到最优运行轨迹;其次识别调度函数的决策变量与自变量,构建线性调度函数;最后采用集合卡尔曼滤波动态更新调度函数的参数,形成适应性调度规则。以锦屏一级水电站及周边风、光电站为实例,研究表明适应性调度规则较常规调度和静态调度,在检验期年均发电量分别提高6.45亿k W·h和0.53亿k W·h,说明论文所提方法可有效追踪最优轨迹的变化来适应变化环境。     

基于集成经验模态分解和最小二乘双支持向量回归机的风速预测算法研究

为有效解决风电大规模并网过程中面临的并网难和弃风等问题,实现可再生能源大规模平滑并网并保证大电网的安全稳定运行,采用集成经验模态分解(ensemble empirical mode decomposition,EEMD)和最小二乘双支持向量回归机(least square twin support vector regression,LSTSVR)算法进行风电场风速预测。分别介绍了LSTSVR、EEMD及自适应变异粒子群算法原理。给出基于EEMD和LSTSVR的风速预测流程,以安徽女儿岭风电场测风声雷达30、70 m处风速采样数据为例,开展基于EEMD和LSTSVR的风速预测算法验证,预测结果误差分析表明:基于EEMD+LSTSVR+自适应变异粒子群算法可以实现风电场风速的高精度预测。     

基于 GBDT 的铁路事故类型预测及成因分析

运用数据挖掘技术进行铁路事故类型预测及成因分析, 对于建立铁路事故预警机制具有重要意义. 为此, 本文提出一种基于梯度提升决策树(Grandient boosting decision tree, GBDT)的铁路事故类型预测及成因分析算法. 针对铁路事故记录数据缺失的问题, 提出一种基于属性分布概率的补全算法, 最大程度保持原有数据分布, 从而降低数据缺失对事故类型预测造成的影响. 针对铁路事故记录数据类别失衡的问题, 提出一种集成的GBDT模型, 完成对事故类型的鲁棒性预测. 在此基础上, 根据GBDT预测模型中特征重要度排序, 实现事故成因分析. 通过在开放数据库上进行实验, 验证了本文模型的有效性.     

Real time flood disaster monitoring based on energy efficient ensemble clustering mechanism in wireless sensor network

In this article, energy efficient ensemble clustering method (EECM) with black widow optimization (EECM-BWO) algorithm is proposed for effective data transmission with the help of real time flood disaster monitoring wireless sensor network (WSN). Initially, unified scalable ensemble clustering algorithm based on ensemble generation and consensus function is proposed for selecting the optimal routing path among the node using BWO algorithm. Then, biologically inspired routing black widow spiders optimization algorithm is proposed to trade off the nodes energy level, self-organization, and self-configuration in the WSN. The simulation is performed using NS2 simulator for validating the performance of the proposed EECM-BWO method. Here, in node, low delay achieves 24.07%, 72.58%, 51.36%, 81.75%, 77.74%, high packet delivery ratio achieves 70.83%, 53.93%, 90.23%, 43.58%, 24.58%, low packet drop attains 77.93%, 72.76%, 61.56%, 51.87%, 34.35%, low energy consumption attains 75.9%, 52.94%, 65.81%, 58%, 41.2% compared with existing energy-efficient clustering approach consolidated game theory as well as dual-cluster-head mode for WSNs energy-aware clustering by cuckoo optimization approach (EECM-COA), energy-aware clustering-based routing using multi-path reliable transmission with routing and control board (EECM-RCB-MRT), adaptive repair algorithm with temporally ordered routing algorithms for flood control strategy (EECM-AR-TORA-FCS), passive multi-hop clustering algorithm (EECM-PMC), dynamic source routing protocol based on genetic algorithm-bacterial foraging optimization (DSR-GA-BFO).     

A simple decision tree-based disturbance monitoring system for VSC-based HVDC transmission link integrating a DFIG wind farm

Fault detection and classification is a key challenge for the protection of High Voltage DC (HVDC) transmission lines. In this paper, the Teager–Kaiser Energy Operator (TKEO) algorithm associated with a decision tree-based fault classi f ier is proposed to detect and classify various DC faults. The Change Identification Filter is applied to the average and differential current components, to detect the first instant of fault occurrence (above threshold) and register a Change Identified Point (CIP). Further, if a CIP is registered for a positive or negative line, only three samples of currents (i.e., CIP and each side of CIP) are sent to the proposed TKEO algorithm, which produces their respective 8 indices through which the, fault can be detected along with its classification. The new approach enables quicker detection allowing utility grids to be restored as soon as possible. This novel approach also reduces computing complexity and the time required to identify faults with classification. The importance and accuracy of the proposed scheme are also thor oughly tested and compared with other methods for various faults on HVDC transmission lines.     

基于生成对抗网络的知识蒸馏数据增强

在图像分类和工业视觉检测过程中,缺陷样本量少导致神经网络分类器训练效率低及检测精度差,直接采用原始的离散标签又无法使网络分类器学习到不同类别间的相似度信息。针对上述问题,在区域丢弃算法的基础上,提出一种基于生成对抗网络的知识蒸馏数据增强算法。使用补丁对丢弃区域进行填补,减少区域丢弃产生的非信息噪声。在补丁生成网络中,保留生成对抗网络的编码器-解码器结构,利用编码器卷积层提取特征,通过解码器对特征图上采样生成补丁。在样本标签生成过程中,采用知识蒸馏算法中的教师-学生训练模式,按照交叉检验方式训练教师模型,根据教师模型生成的软标签对学生模型的训练进行指导,提高学生模型对特征的学习能力。实验结果表明,与区域丢弃算法相比,该算法在CIFAR-100、CIFAR-10数据集图像分类任务上的Top-1 Err、Top-5 Err分别降低3.1、0.8、0.5、0.6个百分点,在汽车转向器轴承数据集语义分割任务上的平均交并比和识别准确率分别提高2.8、2.3个百分点。