基于改进的多块核主元分析的风电机组故障诊断方法

Fault Diagnosis of Wind Turbines Based on Improved Multiblock Kernel Principal Component Analysis

针对风电机组运行工况和数据信息结构复杂、机组故障诊断困难的特点,提出了基于因子分析改进的多块核主元分析(MBKPCA)方法,通过对运行数据信息的深度挖掘,建立了机组数据、变量和运行工况3者之间的关联机制。通过对应分析,明确机组变量与数据的关联性,确定MBKPCA子块数目及实际含义。采用因子分析寻找各子块数据与对应运行过程的相关性,提高MBKPCA诊断精度。结果表明:改进的MBKPCA方法能够及时准确地对风电机组故障进行诊断,具有一定的工程应用价值。

To overcome the difficulty in fault diagnosis of wind turbines due to their complex operation conditions and data information structure, an improved multiblock kernel principal component analysis （MBK- PCA） method was proposed based on factor analysis, so as to establish an association mechanism among the unit data, variables and operating conditions by deeply mining the operation data. Through corresponding analysis, the relationship between unit variables and the data was defined, while the number of MBKPCA sub-blocks and their actual meanings were determined. Finally, the factor analysis was adopted to find out the correlation between the data of various sub-blocks and the process of corresponding motions, thus improving the diagnostic accuracy of MBKPCA. Results show that the improved MBKPCA method can help to make fault diagnosis timely and accurately for wind turbines, which may be applied in actual engineering projects.