非侵入式负荷监测(non-intrusive load monitoring,NILM)是一种无需进入每个用电器内部系统,仅在用户总线入口处安装监测设备的技术.在开展NILM技术研究时,往往需要收集大规模的用户负荷数据来证明所提出方法的普适性,此需求不可避免地带来了繁重的数据收集与整理负担.为克服该挑战,设计了一种结合周期信号频率不变变换(frequency invariant transformation for periodic signals,FIT-PS)原理与时间序列生成对抗网络(time series generative adversarial networks,TimeGAN)的混合模型,记为FIT-PSTimeGAN.针对全球家庭与工业瞬态能量数据集(worldwide household and industry transient energy dataset,WHITED)中的空调、微波炉、吸尘器、冰箱和热水壶5种电器,运用FIT-PS对负荷数据集进行切割和拼接,构建TimeGAN不同状态下的训练集和测试集.评估测试集的效果发现,生成的波形数据与真实数据表现出高度一致性.进一步采用FIT-PS对训练得到的生成数据进行截取和拼接,生成满足测试需求的完整的单负荷波形和多负荷波形.对这些生成的波形与相同状态下的真实数据进行对比,结果显示两者吻合度很高.与自回归模型和生成对抗网络(generative adversarial network,GAN)模型相比,FIT-PS-TimeGAN模型在生成数据的性能方面表现更优.研究结果表明,FIT-PS-TimeGAN混合模型能够有效生成符合标准电器运行规律的波形和场景数据.
为贯彻落实南方电网2016年工作会议精神,推进智能、高效、可靠、绿色的现代化电网建设,开展智能调度建设,进一步对配网方式计划管理模块增加智能分析。方式人员作为网架的指挥官,但因配网方式批复完申请后无法对该申请单进行一个闭环管理,中间环节、闭环环节缺失,大大增加了配网方式对网架工作的时跟踪丢失。对配网计划工作没有一个全面智能分析系统,现在以日、周、月、年的人为分析浪费人力、时间。大大降低工作效率,信息缺失还会对网架有安全隐患。In order to implement the spirit of the 2016 work conference of China Southern Power Grid, promote the construction of an intelligent, efficient, reliable, and green modern power grid, carry out intelligent dispatching construction, and further add intelligent analysis to the distribution network mode plan management module. As the commander of the grid, the personnel can not carry out a closed-loop management of the application form after the application is approved by the distribution network, and the intermediate link and the closed-loop link are missing, which greatly increases the tracking loss of the distribution network mode to the grid work. There is no comprehensive intelligent analysis system for the distribution network planning work, and now the human analysis of the day, week, month, and year is a waste of manpower and time. Greatly reducing the work efficiency, and lack of information will also pose a security risk to the grid.
暂无评论