Abstract: The vibration signals of early faults in rolling bearings are weak and easily submerged by strong background noise. Taking the rolling bearings of the long-shaft deep well pumps in the hydropower station drainage system as the research object, this paper proposes an early fault diagnosis method for rolling bearings fusing spectral kurtosis and 1.5-dimensional spectrum. Firstly, the spectral kurtosis is adopted to adaptively locate the optimal frequency band containing early fault impulses and conduct filtering, so as to preliminarily enhance the fault components. Then, 1.5-dimensional spectrum analysis is performed on the filtered signal, which further highlights the fault characteristic frequencies by virtue of its strong noise suppression capability. Finally, accurate diagnosis of fault types is realized by comparing the peak frequencies extracted from the 1.5-dimensional spectrum with the theoretical fault characteristic frequencies of the bearing. The analysis results of an actual early inner-race fault case of the bearing demonstrate that the proposed method can significantly highlight the weak characteristic components of early faults and greatly improve the reliability and accuracy of early fault feature extraction for rolling bearings.