In order to research start-up pressure wave propagation mechanism and determine pressure wave speed in gelled crude oil pipelines accurately,experiment of Large-scale flow loop was carried out.In the experiment,start-...In order to research start-up pressure wave propagation mechanism and determine pressure wave speed in gelled crude oil pipelines accurately,experiment of Large-scale flow loop was carried out.In the experiment,start-up pressure wave speeds under various operation conditions were measured,and effects of correlative factors on pressure wave were analyzed.The experimental and theoretical analysis shows that thermal shrinkage and structural properties of gelled crude oils are key factors influencing on start-up pressure wave propagation.The quantitative analysis for these effects can be done by using volume expansion coefficient and structural property parameter of gelled crude oil.A new calculation model of pressure wave speed was developed on the basis of Large-scale flow loop experiment and theoretical analysis.展开更多
合环电流评估技术对于配电网馈线合环转供电操作具有重要意义,为了提高主动配电网馈线合环电流计算的准确性,文中从融入源荷数据分布特性及预测的角度,提出基于双重K2算法和概率潮流(double K2 algorithm and probability load flow,DK2...合环电流评估技术对于配电网馈线合环转供电操作具有重要意义,为了提高主动配电网馈线合环电流计算的准确性,文中从融入源荷数据分布特性及预测的角度,提出基于双重K2算法和概率潮流(double K2 algorithm and probability load flow,DK2-PLF)的主动配电网馈线合环电流评估方法。首先,采用基于DK2算法的贝叶斯网络描述源荷相关性样本;其次,利用Cholesky分解方法处理获得的源荷相关性样本,以半不变量法计算主动配电网馈线合环电流的累积概率分布;然后,对主动配电网馈线合环电流从合环成功率和越限程度两方面进行安全性评估;最后,以贵州某城市为算例,对10 kV配电网系统开展馈线合环电流评估研究。得出以下结论:一是从概率密度、累积分布、最大误差三方面比较,相比于K2算法,DK2算法源荷预测值的概率密度、累积分布误差较小,验证了DK2算法的优越性;二是从累积分布、最大误差两方面比较,采用Cholesky分解的半不变量法对比未采用Cholesky分解的半不变量法、蒙特卡洛法,其累积分布误差较小,采用Cholesky分解的半不变量法满足主动配电网馈线合环电流评估要求;三是从合环成功率、合环越限程度两方面比较,采用半不变量法计算的合环电流安全性指标结果与仿真结果偏差在电网经验误差5%范围内,说明基于DK2-PLF的主动配电网馈线合环电流评估方法可为合环辅助决策提供参考。展开更多
当前,知识定义网络赋能AI技术发展,算力网络提供AI所需算力资源,二者逐渐趋于融合,形成了知识定义算力网络(Knowledge Defined Computing Networking,KDCN)。KDCN赋能发展了诸多新型网络应用,如元宇宙、AR/VR、东数西算等,这些新型应用...当前,知识定义网络赋能AI技术发展,算力网络提供AI所需算力资源,二者逐渐趋于融合,形成了知识定义算力网络(Knowledge Defined Computing Networking,KDCN)。KDCN赋能发展了诸多新型网络应用,如元宇宙、AR/VR、东数西算等,这些新型应用对算力资源和网络资源有极大的需求,被称为重击流(Heavy Hitter,HH)。HH流的存在严重加剧了KDCN网络的拥塞情况。针对这一挑战,提出了一种智能流量调度机制,旨在通过深度Q神经网络来解决KDCN中的拥塞问题。相较于离线训练过程,通过流量数据检测与采集、在模型训练和拥塞流调决策之间建立实时闭环,来实现深度Q神经网络模型的在线训练。基于该闭环控制,智能流调模型通过不断学习可以实现持续演化,并用于提供实时决策。实验结果表明,该算法在资源利用率、吞吐量、平均丢包率等方面优于现有方法。展开更多
基金Project(2008B-2901) supported by China National Petroleum Corporation
文摘In order to research start-up pressure wave propagation mechanism and determine pressure wave speed in gelled crude oil pipelines accurately,experiment of Large-scale flow loop was carried out.In the experiment,start-up pressure wave speeds under various operation conditions were measured,and effects of correlative factors on pressure wave were analyzed.The experimental and theoretical analysis shows that thermal shrinkage and structural properties of gelled crude oils are key factors influencing on start-up pressure wave propagation.The quantitative analysis for these effects can be done by using volume expansion coefficient and structural property parameter of gelled crude oil.A new calculation model of pressure wave speed was developed on the basis of Large-scale flow loop experiment and theoretical analysis.
文摘合环电流评估技术对于配电网馈线合环转供电操作具有重要意义,为了提高主动配电网馈线合环电流计算的准确性,文中从融入源荷数据分布特性及预测的角度,提出基于双重K2算法和概率潮流(double K2 algorithm and probability load flow,DK2-PLF)的主动配电网馈线合环电流评估方法。首先,采用基于DK2算法的贝叶斯网络描述源荷相关性样本;其次,利用Cholesky分解方法处理获得的源荷相关性样本,以半不变量法计算主动配电网馈线合环电流的累积概率分布;然后,对主动配电网馈线合环电流从合环成功率和越限程度两方面进行安全性评估;最后,以贵州某城市为算例,对10 kV配电网系统开展馈线合环电流评估研究。得出以下结论:一是从概率密度、累积分布、最大误差三方面比较,相比于K2算法,DK2算法源荷预测值的概率密度、累积分布误差较小,验证了DK2算法的优越性;二是从累积分布、最大误差两方面比较,采用Cholesky分解的半不变量法对比未采用Cholesky分解的半不变量法、蒙特卡洛法,其累积分布误差较小,采用Cholesky分解的半不变量法满足主动配电网馈线合环电流评估要求;三是从合环成功率、合环越限程度两方面比较,采用半不变量法计算的合环电流安全性指标结果与仿真结果偏差在电网经验误差5%范围内,说明基于DK2-PLF的主动配电网馈线合环电流评估方法可为合环辅助决策提供参考。
文摘当前,知识定义网络赋能AI技术发展,算力网络提供AI所需算力资源,二者逐渐趋于融合,形成了知识定义算力网络(Knowledge Defined Computing Networking,KDCN)。KDCN赋能发展了诸多新型网络应用,如元宇宙、AR/VR、东数西算等,这些新型应用对算力资源和网络资源有极大的需求,被称为重击流(Heavy Hitter,HH)。HH流的存在严重加剧了KDCN网络的拥塞情况。针对这一挑战,提出了一种智能流量调度机制,旨在通过深度Q神经网络来解决KDCN中的拥塞问题。相较于离线训练过程,通过流量数据检测与采集、在模型训练和拥塞流调决策之间建立实时闭环,来实现深度Q神经网络模型的在线训练。基于该闭环控制,智能流调模型通过不断学习可以实现持续演化,并用于提供实时决策。实验结果表明,该算法在资源利用率、吞吐量、平均丢包率等方面优于现有方法。
