With the increase of the proportion of new energy and climate-sensitive loads in power system, the uncertainty of system operation is more and more significant, and the complex nonlinear relationship between the power system and the meteorological system is presented, so the N-1 static security assessment of the system is increasingly difficult. To solve this problem, this paper proposes a deep spatiotemporal data-driven model for security assessment of novel power system. A power system security assessment model framework considering local climate is established, and then according to the strong nonlinearity between different physical systems (electrical and meteorological), a power system security assessment method based on deep spatiotemporal data-driven is proposed using data mining technology. In view of the complexity of the deep learning model, based on the prior condition that the climate affects the locality of the power system, the model is simplified and solved using parallel algorithms. An actual power grid is simulated and analyzed. The results show that the weak links of power system security are time-sensitive; the depth model considering local climate can greatly improve the prediction accuracy of safety assessment; model simplification and parallelization can significantly improve learning efficiency.