With the development of power grids and advances in technology, the grid structure is increasingly complex, and it is particularly important to be able to sense the security posture of the grid in a timely and effective manner. Deep learning has made great progress in text, voice, and image in recent years, and it also plays an important role in the field of artificial intelligence. Combining deep learning with the security situational awareness of the grid, a power-based situational awareness based on deep learning is proposed. In the situation understanding stage, from the two aspects of static security and dynamic security of the power grid, a complete set of grid security situation evaluation system is constructed to characterize the running trajectory of the power grid. In the situation prediction stage, a deep learning model is constructed to complete the perception of the security posture of the power grid. Finally, the IEEE39 node system is taken as an example to compare it with BP neural network and RBF neural network prediction model. It is verified that deep learning can effectively sense the security situation of the power grid, and the prediction accuracy is higher than the traditional neural network model.