Abstract: With the large-scale grid integration of renewable energy sources (RES), the power grid operation gradually exhibits a new characteristic of having high-order uncertainty, bring about serious challenges for the system operational security and stability. The traditional methods of model-driven generation scheduling require large amount of computational resources, whereas the widely-concerned Reinforcement Learning (RL) method bears the issues like slow training speed as processing the highly complexed and dimensional grid state information. For this reason, this paper proposes a novel method of Grid Expert Strategy Imitation Learning (GESIL)-based real-time security dispatch. Firstly, a grid model is established based on the graph theory. Secondly, a grid expert strategy considering secured power grid operation and power balance control is proposed. Then, imitation learning is used to combine the grid expert strategy with the proposed model to obtain a GESIL intelligent agent which is used to make specific grid dispatching decisions. A modified IEEE 118 bus system with high RES proportion is employed to compare the proposed GESIL to the traditional scheduling and the RL methods. The results show that the proposed GESIL is able to be more stable and efficient in computing the real-time dispatching decisions of grid operation and power balancing, enhancing the effect and computational speed of the dispatching decision computation.