Abstract: To address the problems of multi-form, non-stationary fluctuations and difficult precise identification of voltage flicker caused by high proportion of distributed photovoltaic (PV) integration in distribution networks, this paper proposes a voltage flicker source identification method combining successive variational mode decomposition and histogram gradient boosting tree algorithm. Firstly, voltage flicker is classified into three typical types, namely sinusoidal wave modulation type flicker, rectangular wave modulation type flicker, and inter-harmonic wave modulation type flicker, based on its causes. For these three types, successive variational mode decomposition is adopted to decompose the sampled voltage into narrowband intrinsic mode functions and extract features. Then, a histogram gradient boosting tree classifier is constructed to identify the voltage flicker sources in distribution networks with high proportion of PV integration. Finally, the simulation results show that the proposed method can effectively decompose different types of flicker signals and accurately identify different voltage flicker sources, with an accuracy rate of 96.9%.