摘要
故障预测是实现电路健康预报的关键技术,对比提出一种基于高斯过程回归(GPR)的方法开展单端初级电感变换器(SEPIC)故障预测研究。通过分析电路关键元器件退化对电路性能的影响,选取输出电压均值作为故障特征参数,依据历史数据建立退化模型,并采用GPR进行递推预测。分析了不同核函数和建模数据规模对预测结果的影响,并与最小二乘法进行对比,验证了所提方法的有效性和准确性。
Fauh prognosis is the key technology to realize health forecast of power converters, a novel fault prediction approach based on Gaussian process regression（GPR） is proposed for single-ended primary inductance converter（SEPIC）. According to analyzing the effect of key components degradation with the performance of the circuit, the average value of output voltage is selected as the fault characteristic parameters.Also ,the degradation model is established on the basis of historical data and GPR method is used for recursive prediction.The influence of different kernel functions and modeling data size on the prediction results is analyzed.Compared with the least square method ,the validity and accuracy of the proposed method are verified.
作者
孙权
王友仁
姜媛媛
邵力为
SUN Quan;WANG You-ren;JIANG Yuan-yuan;SHAO Li-wei(Naujing University of Aeronautics and Astronautics, Nanjing 211106, China)
出处
《电力电子技术》
CSCD
北大核心
2018年第6期17-20,共4页
Power Electronics
基金
国家自然科学基金(61371041)
江苏省普通高校研究生科研创新计划(KYLX_0250)
关键词
功率变换器
故障预测
高斯过程回归
特征参数
power converters
fault prognosis
Gaussian process regression
feature parameter
