论文部分内容阅读
针对传统快速傅里叶变换(FFT)方法在非稳态信号分析上的局限和水电机组稳定性状态分析及故障诊断中过分依赖单一能量特征的不足,提出了集小波分析、模糊理论和径向基函数神经网络(RBFNN)优点于一体的基于复合特征提取的RBFNN故障诊断方法。首先采用小波分析方法对稳定性状态信号进行多频段分解、降噪,提取相对能量特征;运用模糊理论进行稳定性状态对过程参数变化响应的数值分析和量化,提取关系型征兆;然后将这2种特征组合,形成综合反映机组稳定性状态的复合特征向量;最后利用RBFNN诊断出机组的典型故障类型及其严重程度。工程应用结果表明,该方法能够全面准确地提取水电机组稳定性状态特征,在水电机组故障诊断上具有一定的可行性和有效性。
Aiming at the limitations of traditional Fast Fourier Transform (FFT) method in unsteady signal analysis and the lack of reliance on single energy features in hydroelectric generating set stability analysis and fault diagnosis, a set of wavelet analysis, fuzzy theory and radial RBFNN Fault Diagnosis Based on Combined Feature Extraction Based on Advantages of Basis Function Neural Network (RBFNN). First of all, wavelet analysis is used to decompose the stability state signal in multi-band, noise reduction and extract the relative energy characteristics; the fuzzy theory is used to analyze and quantify the response of the steady state to process parameter changes and extract the relational signs; Then the RBFNN is used to diagnose the type of typical fault and the severity of the unit. The engineering application shows that this method can extract the stability characteristics of hydropower units comprehensively and accurately, and is feasible and effective in the fault diagnosis of hydropower units.