Abstract: The ever-increasing uncertainty on both sides of power source and load brings greater challenges to cascading failure risk assessment. Based on the stochastic response surface method and deep forest, a rapid assessment method of cascading failure risk is proposed considering the uncertainty of power source and load. Firstly, an improved stochastic response surface method combining distributed factor（SRSM-DF） is proposed to quickly calculate the probabilistic load flow in the process of cascading line outages.Then, the calculation error judgment method of SRSM-DF is proposed based on the deep forest, which takes into account the speed and accuracy of the probabilistic power flow calculation. Finally, a two-stage high-risk cascading failure screening strategy is constructed. Further considering the load shedding and DC power loss caused by failures, and the linear programming model for obtaining the cascading failure consequences is established to reduce the computational complexity of cascading failure screening.The simulation results of real-life power grid show that the proposed method can quickly screen high-risk cascading failures under uncertain conditions.