摘要
为保证电网安全稳定运行,提高电网防灾减灾和弹性水平,提出了一种基于深度稀疏自编码网络和场景分类器的电网气象故障预警方法。首先,采用主客观权重相结合的动态赋权方法,对气象因子进行初始赋权,以合理表征不同气象因子对电网故障的影响程度。然后,对传统的深度自编码网络增加稀疏性约束条件,以提高网络训练的收敛性,并在深度自编码网络的最后一层增加场景分类器,以提高气象因子与电网故障场景间关联关系的合理性。最后,将带权重的气象因子以及设备因子和环境因子作为深度稀疏自编码网络的输入,利用支持向量机构建多因素耦合的电网气象灾害故障预警模型。采用实际电网故障算例验证了所提方法的有效性。
In order to ensure the safe and stable operation of the power grid and improve the level of disaster prevention,mitigation and resilience of the grid,this paper proposes an early warning method for power grid meteorological faults based on a deep sparse self-encoding network and a scene classifier.First,this paper adopts a dynamic weighting method combining subjective and objective weights to initially weight meteorological factors to reasonably describe the influence of different meteorological factors on power grid faults.Then,a sparsity constraint is added to the traditional deep self-encoding network to improve the convergence of network training,and a scene classifier is added to the last layer of the deep self-encoding network to improve the rationality of the relationship between meteorological factors and power grid fault scenarios.Finally,the meteorological factors which are dynamic weighted,equipment factors and environmental factors are used as the input of a deep sparse self-encoding network,and a support vector machine is used to build a multi-factor coupled grid meteorological disaster fault early warning model.The effectiveness of the method is verified by an actual power grid fault example.
作者
胡谅平
丛伟
徐安馨
魏振
邱吉福
陈明
HU Liangping;CONG Wei;XU Anxin;WEI Zhen;QIU Jifu;CHEN Ming(Key Laboratory for Power System Intelligent Dispatching and Control(Shandong University),Jinan 250061,China;Rizhao Power Supply Company,State Grid Shandong Electric Power Company,Rizhao 276800,China;Qingdao Power Supply Company,State Grid Shandong Electric Power Company,Qingdao 266001,China)
出处
《电力系统保护与控制》
EI
CSCD
北大核心
2022年第20期68-78,共11页
Power System Protection and Control
基金
国家电网公司科技项目资助(52060019001H)。
关键词
电网气象故障
预警方法
动态组合权重
场景分类器
深度稀疏自编码网络
power grid fault caused by meteorology
early warning method
dynamic combination of weights
scene classifier
deep sparse auto-encoder network
