Failure Prediction Based VNF Migration Mechanism for Multimedia Services in Power Grid Substation Monitoring

Abstract

With the development of multimedia Internet of Things (IoT), 5G and mobile edge computing technologies are widely used in smart grids. As a key part in power grid substation monitoring, the delay-sensitive and computation-intensive robot inspection services can be realized by network virtualization and 5G technologies. However, failure protection mechanism in virtual resource management is still a hysteresis problem, which influences the availability of robot inspection system. Therefore, this paper establishes a failure prediction model based on BP neural network for physical nodes and then proposes a deep reinforcement learning based virtual network function migration algorithm according to the failure prediction results. The simulation results show that the presented mechanism can save 10.2% and 13.6% of the cost compared with the heuristic and greedy algorithm.