The power system security is great significant to national security. As one of the main links of power system, power dispatching network is vulnerable to attack by advanced virus software. At present, the physical isolation based on boundary protection is not enough to resist the invasion. Therefore, the power dispatching system network security protection research has very important theoretical value and engineering significance. Based on this, a network state transition algorithm based on Q learning algorithm is constructed to predict the best possible path of network attack. Based on the system loss, attack cost and defense cost, defense return and network state transition cost are introduced. Static economic benefit is taken as the criterion of network state transition. Using evolutionary game theory to construct replication dynamic equation, dynamically reproduce the antagonistic behavior of both attackers and defenders, calculate the dynamic change of economic returns, predict the change of attacking behavior, and determine the best defense strategy. The experimental results show that situational awareness scheme can effectively identify the possible path of network attack and the greatest threat, and help dispatchers make effective decisions.