Dielectric loss angle（DLA）is an important characteristic parameter that reflects the quality of power capacitor. Accurate realization of DLA identification provides reliable data support for early fault diagnosis of capacitor. In case of deviation of sampling frequency or of fluctuation of grid frequency

identification of DLA by using the discrete Fourier transform（DFT）will reduce the measurement accuracy due to the spectrum leakage and the fence effect. The basic way of enhancing the identification accuracy of DLA lies in improving the spectrum leakage effect of the sampling frequency. Therefore

a triangular self-convolution window（TSCW）function is used to perform all phase processing on data and DFT is performed on the new formed data segments respectively to obtain all-phase DFT（apDFT）spectrum analysis results. Furthermore

the spectrum analysis results are corrected by using the improved method for all-phase time-shift phase difference

and the phase difference between voltage and current is extracted from it so to obtain a DLA identification algorithm with higher accuracy. Since the algorithm overcomes the interference of each harmonic to the fundamental wave

it therefore has higher accuracy than the traditional DLA algorithm with windowed interpolation DFT. The simulation and comparison experiment shows that the algorithm can achieve accurately identification of DLA and is less affected by noise

frequency offset and other interference. Finally

the effectiveness and high identification accuracy of the the algorithm are verified by applying it to DLA identification of practical circuits.