多振动信息下的高压断路器机械故障随机森林融合诊断方法

Healthy condition of high voltage circuit breaker (HVCB) has an important impact on the power system. With the development of artificial intelligence, many advanced methods have been applied to fault type identification of HVCBS. At present, most related researches are devoted to improving the feature extraction process or the classification method based on a single sensor to obtain a higher accuracy. However, the improved method can only approach the upper limit determined by data information, ignoring the limited ability of a single information to identify faults. Therefore, this study has proposed a multi-sensor joint decision approach based on random forest. Firstly, under the typical faults condition, the differences of vibration characteristic at the different locations are analyzed. Then, based on a random forest algorithm, a multi-sensor fusion diagnosis process is designed. Finally, based on the HVCB experimental platform, the results of six typical classifiers and random forest fusion method under different sensor combinations are compared to verify that the proposed method can significantly improve fault diagnosis performance and provide new ideas for promoting the application of HVCB fault location.