Variational Inference Based Kernel Dynamic Bayesian Networks for Construction of Prediction Intervals for Industrial Time Series With Incomplete Input

Prediction intervals(PIs)for industrial time series can provide useful guidance for workers.Given that the failure of industrial sensors may cause the missing point in inputs,the existing kernel dynamic Bayesian networks(KDBN),serving as an effective method for PIs construction,suffer from high computational load using the stochastic algorithm for inference.This study proposes a variational inference method for the KDBN for the purpose of fast inference,which avoids the timeconsuming stochastic sampling.The proposed algorithm contains two stages.The first stage involves the inference of the missing inputs by using a local linearization based variational inference,and based on the computed posterior distributions over the missing inputs the second stage sees a Gaussian approximation for probability over the nodes in future time slices.To verify the effectiveness of the proposed method,a synthetic dataset and a practical dataset of generation flow of blast furnace gas(BFG)are employed with different ratios of missing inputs.The experimental results indicate that the proposed method can provide reliable PIs for the generation flow of BFG and it exhibits shorter computing time than the stochastic based one.

Deep reinforcement learning based multi-level dynamic reconfiguration for urban distribution network:a cloud-edge collaboration architecture

With the construction of the power Internet of Things(IoT),communication between smart devices in urban distribution networks has been gradually moving towards high speed,high compatibility,and low latency,which provides reliable support for reconfiguration optimization in urban distribution networks.Thus,this study proposed a deep reinforcement learning based multi-level dynamic reconfiguration method for urban distribution networks in a cloud-edge collaboration architecture to obtain a real-time optimal multi-level dynamic reconfiguration solution.First,the multi-level dynamic reconfiguration method was discussed,which included feeder-,transformer-,and substation-levels.Subsequently,the multi-agent system was combined with the cloud-edge collaboration architecture to build a deep reinforcement learning model for multi-level dynamic reconfiguration in an urban distribution network.The cloud-edge collaboration architecture can effectively support the multi-agent system to conduct“centralized training and decentralized execution”operation modes and improve the learning efficiency of the model.Thereafter,for a multi-agent system,this study adopted a combination of offline and online learning to endow the model with the ability to realize automatic optimization and updation of the strategy.In the offline learning phase,a Q-learning-based multi-agent conservative Q-learning(MACQL)algorithm was proposed to stabilize the learning results and reduce the risk of the next online learning phase.In the online learning phase,a multi-agent deep deterministic policy gradient(MADDPG)algorithm based on policy gradients was proposed to explore the action space and update the experience pool.Finally,the effectiveness of the proposed method was verified through a simulation analysis of a real-world 445-node system.

Method of multi-level recursive and application to nonlinear dynamic deformation forecasting

The time-dependence bilinear mixed-regression deformation model and time-dependence bilinear dynamic system deformation model are established for deformation observation series. According to the multi- level recursive method, the time-dependence parameters are first traced and predicted, and then the dynamic system states. Due to the method considering time-dependence of deformation and having stronger adaptability to time-dependence system, it can improve forecast’s precision. It is very effective for data processing of nonlinear dynamic deformation monitoring to make multi-step forecasting.

Centralized Dynamic Spectrum Allocation in Cognitive Radio Networks Based on Fuzzy Logic and Q-Learning

A novel centralized approach for Dynamic Spectrum Allocation (DSA) in the Cognitive Radio (CR) network is presented in this paper. Instead of giving the solution in terms of formulas modeling network environment such as linear programming or convex optimization, the new approach obtains the capability of iteratively on-line learning environment performance by using Reinforcement Learning (RL) algorithm after observing the variability and uncertainty of the heterogeneous wireless networks. Appropriate decision-making access actions can then be obtained by employing Fuzzy Inference System (FIS) which ensures the strategy being able to explore the possible status and exploit the experiences sufficiently. The new approach considers multi-objective such as spectrum efficiency and fairness between CR Access Points (AP) effectively. By interacting with the environment and accumulating comprehensive advantages, it can achieve the largest long-term reward expected on the desired objectives and implement the best action. Moreover, the present algorithm is relatively simple and does not require complex calculations. Simulation results show that the proposed approach can get better performance with respect to fixed frequency planning scheme or general dynamic spectrum allocation policy.

MEASUREMENT OF WELDING DYNAMIC DISPLACEMENT FIELDS BY ELECTRONIC SPECKLE PATTERN INTERFERENCE

In order to effectively control the stress and distortion which produced in welding process, the dynamic change laws of displacement field is the most important factor. The characteristics of the welding dynamic displacement field is high temperature, high strain velocity, thus ordinary methods such as resistance strain gauge or Moiré method can not be used for the measurement of the zone of high temperature. Speckle interference method has the merits of non-contact, resistance to the disturbance of impure lights, high accuracy of measurement (half of wavelength).The paper represents the measurement of dynamic displacement field of argon-arcspot welding, by which it shows that the method of speckle interference is feasible for the measurement of welding dynamic displacement.

A QoS Mobicast-based dynamic clustering secure multicast scheme for large-scale tracking sensornets

Most of the existing security Mobicast routing protocols are not suitable for the monitoring applications with higher quality of service （QoS） requirement. A QoS dynamic clustering secure multicast scheme （QoS-DCSMS） based on Mobicast and multi-level IxTESLA protocol for large-scale tracking sensornets is presented in this paper. The multicast clusters are dynamically formed according to the real-time status of nodes, and the cluster-head node is responsible for status review and certificating management of cluster nodes to ensure the most optimized QoS and security of multicast in this scheme. Another contribution of this paper is the optimal QoS security authentication algorithm, which analyzes the relationship between the QoS and the level Mofmulti-level oTESLA. Based on the analysis and simulation results, it shows that the influence to the network survival cycle （＇NSC） and real-time communication caused by energy consumption and latency in authentication is acceptable when the optimal QoS security authentication algorithm is satisfied.

Topology inference of uncertain complex dynamical networks and its applications in hidden nodes detection

The topological structure of a complex dynamical network plays a vital role in determining the network＇s evolutionary mecha- nisms and functional behaviors, thus recognizing and inferring the network structure is of both theoretical and practical signif- icance. Although various approaches have been proposed to estimate network topologies, many are not well established to the noisy nature of network dynamics and ubiquity of transmission delay among network individuals. This paper focuses on to- pology inference of uncertain complex dynamical networks. An auxiliary network is constructed and an adaptive scheme is proposed to track topological parameters. It is noteworthy that the considered network model is supposed to contain practical stochastic perturbations, and noisy observations are taken as control inputs of the constructed auxiliary network. In particular, the control technique can be further employed to locate hidden sources （or latent variables） in networks. Numerical examples are provided to illustrate the effectiveness of the proposed scheme. In addition, the impact of coupling strength and coupling delay on identification performance is assessed. The proposed scheme provides engineers with a convenient approach to infer topologies of general complex dynamical networks and locate hidden sources, and the detailed performance evaluation can further facilitate practical circuit design.

Bayesian inference for dynamical systems

Bayesian inference is a common method for conducting parameter estimation for dynamical systems.Despite the prevalent use of Bayesian inference for performing parameter estimation for dynamical systems,there is a need for a formalized and detailed methodology.This paper presents a comprehensive methodology for dynamical system parameter estimation using Bayesian inference and it covers utilizing different distributions,Markov Chain Monte Carlo(MCMC)sampling,obtaining credible intervals for parameters,and prediction intervals for solutions.A logistic growth example is given to illustrate the methodology.

基于多方法的中医体质动态变化因素分析

目的基于一般统计学、Apriori-DEMATEL算法、DoWhy因果推理框架等方法探究中医体质动态变化影响因素。方法对18-60岁人群进行中医体质识别数据的动态采集,包括采集时间、体质类型以及饮食习惯、睡眠习惯、睡眠时间、运动习惯、情绪状态、压力情况、生活环境、工作/生活变故、家庭氛围、出差频次、加班情况11个体质影响因素。采用一般统计分析分析不同类型体质变化对应影响因素的相对百分比,采用Apriori-DEMATEL算法分析饮食习惯等11种体质影响因素与体质变化之间的相关性,采用DoWhy因果推理框架分析饮食习惯、睡眠习惯、睡眠时间、运动习惯、情绪状态、压力情况之间的因果关系。探讨构成类型转换变化因素的频率,确定构成类型动态变化的关键影响因素。结果经过预处理后形成13536条有效数据,基于Apriori-DEMATEL算法将因素划分成饮食习惯、睡眠习惯、睡眠时间、运动习惯、情绪状态和压力情况这6个原因因素及居住环境、工作/生活变故、家庭氛围、出差频次和加班情况这5个结果因素。结合一般统计学分析发现,在原因因素中饮食习惯、睡眠习惯、睡眠时间和压力情况这4个因素的变化对其他因素的影响程度大,在体质调摄过程中,应对上述4个因素重点关注,保持体质的平衡;在5个结果因素里面,工作/生活变故和家庭氛围这两个因素的数值绝对值较大,说明这两个因素易受其他因素的影响。饮食习惯、睡眠习惯、睡眠时间、运动习惯、情绪状态和压力情况这六个因素变化的中心度较高,说明这六个因素的变化是体质发生改变的重要性因素。根据各因素对应体质变化频数统计发现以上六个因素的变化在气虚质、平和质和特禀质间的体质转变频数占比较大,说明这六个因素的改变对这三种体质类型的变化发挥重要作用。结合Apriori-DEMATEL算法和DoWhy因果推理框架分析的结果,推断饮食习惯和睡眠时间通过影响其他因素的变化间接导致体质变化。结论本文从动态数据和多种分析方法入手,探讨了中医体质动态变化的影响因素,结果表明饮食习惯、睡眠习惯、睡眠时间、运动习惯、情绪状态和压力情况的变化对气虚质、平和质和特秉质的转化有较大影响。在日常生活中要注意这六个因素的变化,并制定相应的改善方案,减少转化为偏体质的概率。本研究也为中医体质类型动态变化的影响因素分析提供了数据支持和客观化分析参考。

水声通信信道估计技术的研究进展

水声信道估计是水声通信的关键技术,其估计性能直接影响通信系统的可靠性和传输速率。介绍了水声通信技术发展的不同阶段、水声信道相关特性以及其对水声通信系统的影响,对现有的水声信道估计技术进行分类,讨论了各自的优势与不足,最后分析了水声信道估计技术面临的问题并提出了解决思路。

动态桌面推演模式下“签派资源管理”模块训练脚本设计研究

为提高签派员的胜任力,减少航空公司因签派员人为原因导致的不安全事件,通过梳理传统“签派资源管理”模块训练存在的不足,以波音B737-800飞机执行“广州—乌鲁木齐”航线为例,融入动态桌面推演理念,完成训练脚本设计,弥补签派员胜任力短板。研究建议:设计训练脚本应从深入一线调研、融入行业热点和持续优化教案三个方面展开,持续做好训练脚本研发、测试、实施和优化的闭环管理。

基于力学机理的贝叶斯动态线性岩质边坡变形预测模型研究

考虑岩质边坡变形的力学机理,将边坡变形分解为弹塑性损伤变形、粘弹性变形、周期波动及观测误差等分量,以Bayesian动态线性模型理论为基础,建立了基于力学机理的边坡变形动态演化预测模型。模型以边坡岩体力学参数为系统控制变量,应用Bayesian推断,通过观测数据及相关信息,估计其动态演化规律,并以此预测边坡变形。通过对某水电站边坡位移监测数据分析表明:监测点处边坡变形主要为弹塑性损伤蠕变,监测初期,岩体迅速损伤,变形15天后进入加速蠕变阶段,与现场巡查结果一致。

基于动态不确定因果图的航天器故障诊断方法

针对航天器智能化故障诊断的问题,基于动态不确定因果图(Dynamic Uncertain Causality Graph,DUCG)构建诊断模型,克服了基于规则的方法、数据驱动方法存在的诊断正确率低、数据依赖程度高、可解释性差等问题。DUCG基于领域专家的经验知识、以图形化的方式表达航天器遥测参数与可能的故障之间的不确定性知识,不依赖于已有的故障数据,具有诊断正确率高、可解释性强等特征。使用DUCG构建包含42个故障、129个遥测参数的诊断模型,试验结果表明模型的准确率为100%。

Evidence of mantle-rooted fluids and multi-level circulation ore-forming dynamics:A case study from the Xiadian gold deposit,Shandong Province,China

Metallogensis of the Xiadian gold deposit in Shandong Province has been a question under dispute for a long time. There are many points such as metamorphic hydrothermal, magamatic hydrothermal and meteoric water. Detailed study shows that mantle-rooted fluids were involved in the ore-forming processes. Evidence for this argumentation comes from: (1) discordogenic fault; (2) intersecting and accompanying of basic veins and lodes; (3) geochemistry of stable isotopes; (4) geochemistry of fluid inclusions; and (5) multi-level circulation and exchanging of mantle-rooted fluids. Based on the characteristics of the circulation system of mantle-rooted fluids and its close relation to magmatic hydrothermal fluids and meteoric water, ore-bearing fluids are divided into three subsystems: (1) C-H-O-rich fluid circulation subsystem in mantle, (2) Si-rich fluid circulation subsystem in the middle and lower crust; and (3) S-rich fluid circulation subsystem in shallow and surface crust. Ore-forming functions of these subsystems are controlled respectively by their different geodynamic settings.

基于动态采样对偶可变形网络的实时视频实例分割

为了更好地利用视频帧中蕴含的时间信息,提升视频实例分割的推理速度,提出动态采样对偶可变形网络(DSDDN). DSDDN使用动态采样策略,根据前、后帧的相似性调整采样策略.对于相似性高的帧,该方法跳过当前帧的推理过程,仅使用前帧分割进行简单迁移计算.对于相似性低的帧,该方法动态聚合时间跨度更大的视频帧作为输入,对当前帧进行信息增强.在Transformer结构里,该方法额外使用2个可变形操作,避免基于注意力的方法中的指数级计算量.提供精心设计的追踪头和损失函数,优化复杂的网络.在YouTube-VIS数据集上获得了39.1%的平均推理精度与40.2帧/s的推理速度,验证了提出的方法能够在实时视频分割任务上取得精度与推理速度的良好平衡.

基于全局图推理与改进三维动态卷积的鱼类摄食行为分析

本研究提出一种基于时间动作检测的轻量化视频分类网络,旨在解决水产智能化养殖中饵料的投喂不均和水体污染等问题,提高投喂准确性和效率。该网络以ResNet 3D为基础,引入深度可分离卷积模块和三维动态卷积模块,以降低模型规模和参数量;同时采用图卷积全局推理模块和稠密卷积模块构建区域和全局关系,增强网络深层特征的表达,提高网络分类准确率。经试验验证,该模型检测准确率可达96.70%,相较变分自动编码器卷积网络和3D ResNet-GloRe网络,其准确率分别提高7.7个百分点和4.4个百分点;同时,该模型的参数量和计算量也明显降低,分别为1.10 M和3.87 G。研究结果表明,该基于时间动作检测的轻量化视频分类网络可以有效提高水产养殖中饵料的智能化投喂的准确性和效率,减少饵料投喂不均以及水体污染等问题,具有较高的应用价值。

Multi-objective Dynamic Reconfiguration for Urban Distribution Network Considering Multi-level Switching Modes

The increasing integration of photovoltaic generators(PVGs) and the uneven economic development in different regions may cause the unbalanced spatial-temporal distribution of load demands in an urban distribution network(UDN). This may lead to undesired consequences, including PVG curtailment, load shedding, and equipment inefficiency, etc. Global dynamic reconfiguration provides a promising method to solve those challenges. However, the power flow transfer capabilities for different kinds of switches are diverse, and the willingness of distribution system operators(DSOs) to select them is also different. In this paper, we formulate a multi-objective dynamic reconfiguration optimization model suitable for multi-level switching modes to minimize the operation cost, load imbalance, and the PVG curtailment. The multi-level switching includes feeder-level switching, transformer-level switching, and substation-level switching. A novel load balancing index is devised to quantify the global load balancing degree at different levels. Then, a stochastic programming model based on selected scenarios is established to address the uncertainties of PVGs and loads. Afterward, the fuzzy c-means(FCMs) clustering is applied to divide the time periods of reconfiguration. Furthermore, the modified binary particle swarm optimization(BPSO)and Cplex solver are combined to solve the proposed mixed-integer second-order cone programming(MISOCP) model. Numerical results based on the 148-node and 297-node systems are obtained to validate the effectiveness of the proposed method.

针对动物领导关系的可变寻正时滞传递熵

为改进可变时滞传递熵(VL传递熵),提出一种可变寻正时滞传递熵(VPL传递熵)。根据提出的时滞寻正设想,重新确定用来矫正DTW时滞序列的最佳固定时滞,用经矫正后的最佳时滞序列对时序进行“扭曲”,进行因果推断。在两个公开的动物运动数据集中进行领导关系因果推断,实验结果表明,VPL传递熵因果推断的准确度比VL传递熵提升了75%至100%。所提方法在动物群体领导关系发现上优于VL传递熵。

强化学习框架中因果推断研究进展

因果推断在科学领域备受关注。近年来,因果关系的确定方法有所创新。强化学习作为一种机器学习方法,主要关注智能体如何在环境中采取行动,以最大化累积奖励。将因果推断方法嵌套在强化学习框架中的思想是因果推断领域以及强化学习方法论中重要的学术进展。基于此,首先,梳理了深度强化学习算法的背景和发展,介绍了基于值函数、基于策略梯度和基于模型的3类强化学习算法框架,以及与因果推断相结合的方向;其次,从5个技术应用角度,对强化学习思想在因果推断和因果识别中的应用研究进行了综述;最后,强调了强化学习框架中因果推断的数据驱动效率、稳定性及应用研究的必要性,并对未来的研究方向进行了展望。

基于动态图表示的设备故障推理预测方法

高效的设备运维可以保障设备的正常运行。然而,随着设备复杂程度越来越高,设备的维护和故障排查的复杂度和难度也不断增加。因此,人工方式越来越不能满足智能化设备的运维需要。智能运维将人工智能等新兴技术运用于运维过程,可以作为设备运维的有力支撑。但现有的很多方法依旧存在着未考虑动态性等不足。针对上述问题,提出了一种基于动态知识图谱表示学习的设备故障推理预测方法,用于预测目标设备是否与故障设备存在潜在的关联。该方法结合动态知识图谱表示学习和图表示推理模型,可以利用实时数据更新图网络,并运用图表示推理模型对新的故障数据进行推理。首先,使用动态知识图谱来表示设备运维数据,记录设备随时间的演化过程,从而有效地表达设备之间关系的动态变化性;然后,通过表示学习获得动态知识图谱中源故障设备和目标设备的时间感知表示;最后,将时间感知表示作为输入进行故障推理预测,判断设备之间是否存在潜在的关联。预测结果可以辅助运维人员解决相应的设备故障问题。在多个数据集上进行了实验,验证了所提方法的有效性。