图5和表1表示不同故障诊断方法运行50次的精度识别结果，从图5中可以看

Fig. 5 and Table 1 show the accuracy recognition results of different fault diagnosis methods running 50 times. It can be seen from Fig. 5 that when m = 3, 4, 5, 6 this method FFCMPE recognition rate is the highest. It can be further seen from Table 1 that when m = 3, 4, 5, and 6, the recognition ability of FFCMPE is better than the five methods of FFMPE, SMPE, FFMFE, SCMPE, and SMFE. Among them, when m = 4, the highest accuracy recognition rate of FFCMPE reaches 96.76%, FFMPE, SCMPE and SMPE are better recognition degrees, SMFE and FFMFE are the worst recognition accuracy, and FFCMPE is more than 5 kinds of FFMPE, SMWPE, FFMFE and SMFE The recognition accuracy of the method is 8.54% -10.54%, 12.53% -15.53%, 17.08% -22.69%, 26.75% -30.75%, 34.84% -39.17%, respectively. The test results verify the advantages of FFCMPE in the accuracy of OLTC recognition and the advantages of feature extraction. At the same time, the standard deviation (STD) of the five methods of FFMPE, SMPE, FFMFE, SCMPE and SMFE is significantly greater than the standard deviation of MWPE multi-feature fusion method. This shows that the calculation results of this method have good stability.

图5和表1表示不同故障诊断方法运行50次的精度识别结果，从图5中可以看出，当m=3,4,5,6本文方法FFCMPE的识别率是最高的。从表1中可以进一步看出m=3、4、5、6时，FFCMPE的识别能力优于FFMPE、SMPE、FFMFE、SCMPE和SMFE5种方法。其中当m=4时，FFCMPE的最高精度识别率达到96.76%，FFMPE、SCMPE和SMPE为较好的识别度，SMFE和FFMFE为最差的识别精度，且FFCMPE比FFMPE、SMWPE、FFMFE和SMFE5种方法分别识别正确率高出8.54%-10.54%，12.53%-15.53%，17.08%-22.69%，26.75%-30.75%，34.84%-39.17%。试验结果验证了FFCMPE在OLTC识别精度方面的优势，以及在特征提取方面的优势，同时FFMPE、SMPE、FFMFE、SCMPE和SMFE5种方法的标准偏差(STD)明显大于MWPE多特征融合方法的标准差，这表明该方法的计算结果具有较好的稳定性。

Figure 5 and table 1 show the accuracy identification results of different fault diagnosis methods running for 50 times. It can be seen from Figure 5 that when m = 3,4,5,6, the identification rate of ffcmpe method in this paper is the highest. It can be further seen from table 1 that when m = 3, 4, 5 and 6, ffcmpe has better recognition ability than ffmpe, SMPE, ffmfe, scmpe and smfe5 methods. When m = 4, the highest recognition accuracy of ffcmpe is 96.76%, ffmpe, scmpe and SMPE are better, SMFE and ffmfe are the worst, and ffcmpe is 8.54% - 10.54%, 12.53% - 15.53%, 17.08% - 22.69%, 26.75% - 30.75%, 34.84% - 39.17% higher than ffmpe, smwpe, ffmfe and SMFE respectively. The experimental results verify the advantages of ffcmpe in OLTC recognition accuracy and feature extraction. At the same time, the standard deviation (STD) of ffmpe, SMPE, ffmfe, scmpe and smfe5 methods is significantly greater than that of mwpe multi feature fusion method, which shows that the calculation results of this method have good stability.<br>