Due to the prospects for development the ubiquitous power Internet of things (IoT), the edge computing is becoming one of the key technologies, which can meet the low latency, high security and data privacy requirements that is essential for the smart grids. By taking advantages of short distance to the power terminals and edge computing devices, massive data can be collected with low delay and deep learning algorithm can be employed to provide varieties adaptive real time control, such as real time power prices prediction, load balancing in intelligent energy system and so on. In this paper, several algorithms are introduced to illustration the novel approaches to provide the efficiency and security for edge computing assisted smart grids.