基于雨雾分离处理和多尺度网络的图像去雨方法

雨带来的雨条纹和雨雾会降低户外拍摄图像的质量,为了去除雨雾对图像的影响,提出了一种基于雨雾分离处理和多尺度卷积神经网络的图像去雨方法。首先利用导向滤波将雨线和图像细节信息提取到高频层,雨雾和背景信息则分离到低频层;然后构建多尺度卷积神经网络来去除高频层中的雨线,网络中融入多个稠密连接模块以提升特征提取的准确性;其次构建多层特征融合的轻量级去雾网络来去除低频层中的雨雾,采用参数一体化结构避免了估计多个大气散射模型参数导致的次优解;最后再结合处理后的高低频结果还原出清晰图像。在多个合成的雨雾数据集以及真实自然场景图像上进行测试,定性和定量结果表明,提出的方法在去除雨雾影响的同时较好地保留了色彩信息,和近年的算法相比,图像结构相似性约提升了0.02~0.08,图像峰值信噪比约提升了0.2~3.5 dB。     

Measuring code maintainability with deep neural networks

The maintainability of source code is a key quality characteristic for software quality. Many approaches have been proposed to quantitatively measure code maintainability. Such approaches rely heavily on code metrics, e.g., the number of Lines of Code and McCabe’s Cyclomatic Complexity. The employed code metrics are essentially statistics regarding code elements, e.g., the numbers of tokens, lines, references, and branch statements. However, natural language in source code, especially identifiers, is rarely exploited by such approaches. As a result, replacing meaningful identifiers with nonsense tokens would not significantly influence their outputs, although the replacement should have significantly reduced code maintainability. To this end, in this paper, we propose a novel approach (called DeepM) to measure code maintainability by exploiting the lexical semantics of text in source code. DeepM leverages deep learning techniques (e.g., LSTM and attention mechanism) to exploit these lexical semantics in measuring code maintainability. Another key rationale of DeepM is that measuring code maintainability is complex and often far beyond the capabilities of statistics or simple heuristics. Consequently, DeepM leverages deep learning techniques to automatically select useful features from complex and lengthy inputs and to construct a complex mapping (rather than simple heuristics) from the input to the output (code maintainability index). DeepM is evaluated on a manually-assessed dataset. The evaluation results suggest that DeepM is accurate, and it generates the same rankings of code maintainability as those of experienced programmers on 87.5% of manually ranked pairs of Java classes.     

基于边权表示学习的乌克兰互联网变化感知

2022年2月至7月期间,网络攻防频繁发生,乌克兰部分地区间接受到断网影响,其境内和边界网络连通情况发生了改变,通过对时间变化拐点的感知和变化区域的分析可敏锐捕捉网络异常变化的细节,为网络威胁预警、网络关键基础设施风险评估等提供重要的信息支持。本文针对2022年2月至7月期间乌克兰自治系统的路由变化进行研究,构建乌克兰自治系统动态网络拓扑,介绍动态网络分析方法,监测乌克兰网络变化,分析断网的时间、地点,并感知断网后网络发生异常的地区。通过与现实事件的对比分析,映射出互联网结构变化感知结果与实体军事冲突的相关性,为从互联网空间观测复杂战场态势提供了信息支撑。     

Adaptive prescribed time control for quadrotor formation with stochastic links failure

The formation control for multiple quadrotors subject to maintaining the formation configuration and collision avoidance in the situation of stochastic links failure is investigated in this paper. First, the distributed formation controller is designed, the position controller is developed to manage the desired formation of position, and the attitude controller is developed to control the translation and rotation movements of the quadrotor. Then, in order to avoid the collisions between multiple quadrotors and the obstacles, a potential energy function method is introduced into the quadrotor formation control combined with the nest adaptive control. Inspired by the design of event trigger controller, a communication compensation controller is designed to ensure the stability of quadrotor formation under the condition of random communication interruption and recovery. Moreover, a prescribed time function is designed, which means the convergence time of the formation system can be set in advance. The prescribed time stability of the formation control system is proved by Lyapunov theory. Finally, the simulation results verify the effectiveness and superiority of this method.     

基于改进深度Q网络算法的多园区综合能源系统能量管理方法

多园区综合能源系统可通过多能互补互济显著提升运行经济性,然而园区之间的复杂互动、多能耦合决策会给多园区综合能源系统的能量管理带来决策空间庞大、算法难以收敛等挑战性问题。为解决上述问题,提出了一种基于改进深度Q网络(modified deep Q network,MDQN)算法的多园区综合能源系统能量管理方法。首先,采用独立于园区的外部气象数据、历史互动功率数据,构建了基于长短期记忆(long short-term memory,LSTM)深度网络的各园区综合能源系统外部互动环境等值模型,降低了强化学习奖励函数的计算复杂度;其次,提出一种基于k优先采样策略的MDQN算法,用k-优先采样策略来代替ε贪心策略,克服了大规模动作空间中探索效率低下的问题;最后,在含3个园区综合能源系统的算例中进行验证,结果表明MDQN算法相比原DQN算法具有更好的收敛性与稳定性,同时可以提升园区经济效益达29.16%。     

基于时域有限积分技术的四级串联快脉冲直线型变压器驱动源电磁特性

快脉冲直线型变压器驱动源（FLTD）感应腔内部结构和媒质分布复杂,支路放电时快速上升沿脉冲将使空间离散更为繁密,导致计算量剧增。该文利用时域有限积分理论首次研究4级串联共用、独立腔体结构MA级FLTD在支路放电时的电磁场分布规律。使用场路耦合将电路中电容元件的放电电流作为激励耦合至三维场模型,使用非均匀网格对FLTD初级感应腔内的关键部件剖分网格进行加密处理,使用表面阻抗边界条件描述良导体以降低求解自由度数量,使用基于等效原理的区域分解算法将4级串联FLTD划分为24个不同的求解子区域。结果表明,时域有限积分技术可有效分析FLTD初级腔体内脉冲放电瞬态过程,两种感应腔结构的输出性能和电磁场分布基本一致,4级串联共用腔体结构FLTD设计方案可行。     

适于深层次CNN的配电网过电压数据集建立方法

随着人工智能算法的快速发展,在配网过电压识别中层数较少的卷积神经网络已得到应用。深层次网络有较高的识别率,但需要大量数据样本,目前已有数据集数据量不足,不能满足深层级网络训练所需。为此提出了一种满足深层次网络训练所需的配电网过电压数据集建立方法。首先利用电磁暂态仿真软件EMTPworks仿真10 kV配网5种典型的过电压并且编辑相应的JavaScript脚本,通过改变故障初相角、过渡电阻、线路长度等参数产生16 272个数据。然后将三相过电压一维信号进行连续小波变换,得到相应二维时频图,并根据原始信号的特征自动标记二维时频图,从而建立了完整的配电网过电压数据集。最后利用卷积神经网络（CNN）对5类过电压信号数据的有效性进行验证。结果表明,构造的数据集数据规模大,有效性高,能够满足深层次网络需要。     

多联产鼓泡流化床内气固流动特性数值模拟研究

为了深入探究流化床内的气固流动规律,以浙江大学1 MW循环流化床热电气多联产试验台为研究对象,使用开源软件MFiX对多联产系统热解炉在不同热解温度（500、600、700 ℃）时稳态的气体组分分布和炉内固体颗粒相体积分数的变化规律进行了数值模拟。模拟过程使用计算流体力学中被广泛应用于流化床数值模拟的欧拉双流体模型和Gidaspow曳力模型,采用Fortran语言编写热解反应子程序,基于OpenMpI技术实现了三维流化床热解炉的并行化模拟。结果表明:数值模拟结果能准确预测稳态运行时炉膛出口的气体组成以及炉内轴向固相体积分数呈近“S”型分布的环-核结构现象。