The invention relates to a probabilistic power flow analysis method for a photovoltaic grid-connected system based on random variable state timing simulation. The method comprises, establishing, basedon the analysis of actual photovoltaic power plant data, a BP neural network model for short-term prediction of photovoltaic generation power; timing simulating the running status of a line in the power grid by a Monte Carlo method, researching the probabilistic power flow of the system by comprehensively considering the uncertainty of photovoltaic generation power and the operating status of theline; and researching an example based on an IEEE 30-node system, wherein the case analysis result shows that a four-input one-output photovoltaic generation power prediction method can be established for actual prediction. In addition, based on the analysis of probabilistic power flow such as node voltage and line power, the necessity of probabilistic power flow research for the photovoltaic grid-connected system is verified.