Based on the fuzzy C-means（FCM） clustering algorithm， a new method for probabilistic voltage stability evaluation considering bus load correlation and load forecasting uncertainty is presented in this paper via adopting the multi-dimension correlation sampling technique. The probabilistic voltage stability of power system is evaluated by the statistical information of system maximum loadability and system participations， which are obtained by repeated load level sampling as well as afterward continuation load flow calculation and modal analysis near the voltage collapse point， thus the voltage instability probability at a certain load level and the area prone to voltage instability from the viewpoint of probability are obtained. The proposed method is tested in IEEE 57-bus system to show its feasibility and validity， The effects of the number of fuzzy clustering， bus load correlation and uncertainty on voltage stability are investigated.