摘要
受机械因素、电气因素和水力因素的耦合作用,水电机组振动信号表现出明显的非线性、非平稳特性,是一类典型的分形信号。基于此,提出了一种基于多重分形与二进制引力搜索算法优化概率神经网络(PNN)的水电机组信号特征提取方法,通过多重分形分析方法提取信号特征,使用二进制引力搜索算法进行特征降维,将降维后的特征向量输入PNN识别,并与降维前的特征向量以及使用EMD模糊熵提取的特征向量进行对比。结果表明,所提方法能够准确地提取机组振动信号特征,提高机组状态识别准确率,为提高电站的安全、稳定运行提供理论依据。
Affected by the coupling of mechanical,electrical and hydraulic factors,the vibration signal of hydropower unit exhibits obvious nonlinear and non-stationary characteristics,which is a typical fractal signal.This paper proposes a signal feature extraction method for hydropower units based on multi-fractal and binary gravity search algorithm to optimize the probabilistic neural network(PNN).The signal feature is extracted by the multi-fractal analysis method,and the binary gravity search algorithm is used to reduce the dimension of the feature.The vector input is recognized by PNN and compared with the feature vector before dimension reduction and the feature vector is extracted by using EMD fuzzy entropy.The results show that the method proposed in this paper can accurately extract the characteristics of unit vibration signals,improve the accuracy of unit status identification,and provide a theoretical basis for improving the safe and stable operation of the power station.
作者
安宇晨
郑阳
陈启卷
席慧
闫懂林
游仕豪
AN Yu-chen;ZHENG Yang;CHEN Qi-juan;XI Hui;YAN Dong-lin;YOU Shi-hao(School of Power and Mechanical Engineering,Wuhan University,Wuhan 430072,China)
出处
《中国农村水利水电》
北大核心
2021年第10期145-150,共6页
China Rural Water and Hydropower
基金
国家自然科学基金项目(52009096)。
关键词
水电机组
多重分形分析
二进制引力搜索算法
概率神经网络
状态识别
hydropower unit
multi-fractal analysis
binary gravity search algorithm
probabilistic neural network
state identification
