中文摘要：

变压器振动信号频谱具有稀疏性,传统的信号分析方法需要计算整个频率范围内的频谱成分,计算速度慢。稀疏快速傅里叶变换（sparse fast Fourier transform,SFFT）算法只计算变压器振动信号的主要频谱成分,利用窗函数过滤信号,然后散列傅里叶系数,最后进行定位与估值运算,能快速的计算出信号频谱中k个拥有最大值的傅里叶系数。该算法结构简单,运行时间相对于信号长度n呈亚线性。通过分析变压器油箱的实际振动信号,验证了SFFT算法较之FFT算法运行速度快,非常适合振动信号的在线频谱分析。

英文摘要：

The frequency spectrum of the transformer vibration signal of is sparse, and to the traditional signal analysis methods, frequency components in the whole frequency range need to be calculated, so the calculation speed is slow. To the sparse fast Fourier transform （SFFT） algorithm, only the main frequency components of transformer vibration signal are calculated. First, SFFT algorithmutilizes window function to filter vibration signal. Then, after the Fourier coefficients being hashed, the largest coefficients ofthe Fourier Transform of vibration signal can be estimated by location and estimation methods. SFFT algorithm with sub linear runtime in the signal size has a simple structure The analysis result of vibration signal of transformer oil tankin this paper has verified the faster performance of the SFFT algorithm than FFT algorithm in on-line spectrum analysis.