首页|期刊导航|苏州科技大学学报(自然科学版)|基于多传感器图Transformer的高频电力电子变换器故障诊断研究

基于多传感器图Transformer的高频电力电子变换器故障诊断研究OA

Fault diagnosis of high-frequency power electronic converters based on multi-sensor graph transformer

中文摘要英文摘要

电力电子变换器的故障诊断对于维护变换器的正常运行至关重要.针对当前电力电子变换器故障诊断方法在准确度方面的不足,研究提出了一种基于多传感器与图Transformer的变换器故障诊断模型.研究采用电流传感器等多个传感器对变换器进行监测,并采用动态时间规整与连续小波变换进行数据处理,最后通过构建图Transformer模块进行特征提取与故障诊断.在仿真实验中,研究提出模型对于电容故障、电感故障问题的检测准确率分别为99.87%和99.91%,对于二极管短路问题的检测准确率为99.67%,均高于对比模型,具有更好的诊断性能.实验结果表明,研究提出模型具有更高的故障诊断准确度,能够更精确地完成高频电力电子变换器故障诊断任务.

Fault diagnosis of power electronic converters is crucial for maintaining their normal operation.In view of the deficiency in accuracy of the current fault diagnosis methods for power electronic converters,a fault diagnosis model for converters based on multi-sensors and graph Transformer is proposed.We employed multiple sensors to monitor the converter,and adopted dynamic time wrapping and continuous wavelet transform for data processing.Fi-nally,a graph Transformer module was constructed for feature extraction and fault diagnosis.In the simulation ex-periment,the detection accuracy rates of the proposed model for capacitor faults and inductor faults were 99.87%and 99.91%respectively,and the detection accuracy rate for diode short circuits was 99.67%,both of which were higher than those of the comparison model,demonstrating better diagnostic performance.The experimental results show that the model proposed has higher accuracy in fault diagnosis and can complete the fault diagnosis task of high-frequency power electronic converters precisely.

刘宇环;金占双

合肥财经职业学院 人工智能与制造学院,安徽 合肥 230601河北北方学院 理学院,河北 张家口 075000

信息技术与安全科学

电力电子变换器图卷积Transformer小波变换传感器

power electronic convertergraph convolutionTransformerwavelet transformsensor

《苏州科技大学学报(自然科学版)》 2026 (3)

83-88,6

安徽省高等学校自然科学研究重点项目(2023AH052565)

10.12084/j.issn.2096-3289.2026.03.010

评论