Fractured tight clastic rock is an important hydrocarbon reservoir，and identification and prediction of its fracture development section is important. Research indicates that the principal component analysis method is effective in fracture identification. In order to further improve the defect of principal component analysis in dealing with data non-linear structure，the method is projected into the Hilbert space，to form the kernel principal component analysis（kernel PCA）utilizing the kernel function，which effectively improves the ability of non-linear data analysis and the accuracy of data extraction and identification. The effective fractures of the tight clastic reservoir of Shanxi Formation in south Qinshui Basin are identified. The cumulative contribution rate of the principal component 1 and 2 is increased to 94.800%，which is 6.240% higher than that of the traditional principal component analysis（PCA）. Using the optimized kernel PCA to identify the effective fractures can more effectively distinguish the fractured stratum and the non-fractured stratum of the dense clastic rock，and can further identify the unfilled，the half-filled，and the full-filled fractured formation，which improves the identification accuracy of the effective fracture in dense clastic rock reservoirs.