Abstract: The N-1 contingency screening is a significant tool to secure power systems. To consider the impact of increasing uncertainty on power system operation, the N-1 contingency screening should analyze the system states with a large number of uncertain scenarios and contingency screening. It makes the calculation of the N-1 contingency screening highly challenging. An efficient method based on deep learning techniques was proposed in this paper to solve this problem, from the flowing two aspects: feature vector construction and learning strategy design. This paper constructed feature vectors representing the variations of source, load, and topology in power systems, and thus established the deep neural network-based DC power flow model. Considering that there are some outliers in the power flow of the N-1 contingency screening, a deep learning strategy focusing on the data-preprocessing method and activation function was proposed for the N-1 contingency screening. After training, the deep neural network-based DC power flow model is capable to solve the power flow of different unseen uncertainty and contingency scenarios. Besides, it can map the results of all the needed analyzed scenarios by matrix multiplier, which can accelerate the calculation speed of N-1 contingency screening with uncertainty scenarios. Finally, the effectiveness of the proposed method was verified in the modified IEEE 39 and IEEE 118 systems.