摘要
针对水电机组空蚀信号非平稳和非线性的特点,提出一种基于经验模态分解-BP神经网络(EMDBPNN)的空蚀故障混合特征提取与分类方法。首先对空蚀信号进行经验模态分解,得到一系列的本征模态函数(IMFs),提取各IMFs分量的能量特征和奇异值特征,同时提取常规的时域和频域特征,构建混合特征向量;然后将此向量作为神经网络的输入,对水电机组空载工况、导叶30%开度和满负荷运行等三种工况下的空蚀数据进行识别分类。试验结果显示,该方法能够对水电机组空蚀故障进行准确诊断,具有较强的工程应用价值。
Due to the characteristic of non-stationary and non-linear hydropower units cavitation signals,a hybrid feature extraction and classification method based empirical mode decomposition（EMD）and back propagation neural network（BPNN）for cavitation fault was proposed in this paper.Firstly,cavitation signals were decomposed into several intrinsic mode functions（IMFs）.Then,energy features and singular value features were extracted from IMFs.Meanwhile,some features both in time domain and frequency domain were extracted for the construction of mixed features vectors.Finally,these vectors were taken as the inputs of neural network.A case study including the classification of cavitation fault under conditions of no-loading,30% guide vane opening and full-loading was used to verify the effectiveness of the proposed method.Experimental results show that the method can classify the cavitation signals effectively and it has strong engineering application value.
出处
《水电能源科学》
北大核心
2018年第3期157-160,共4页
Water Resources and Power
基金
国家重点研发计划(2016YFC0401905)
国家自然科学基金项目(51079057)
国家电网公司科技项目(SGXY-2017J02-087)
