With the large-scale development of renewable energy and the high proportion of massive terminals connected to the grid, the network load in the next generation smart grid will be further intensified, which brings unprecedented and great challenges to the power sensing network for real-time data collection and processing, and whole-domain information monitoring. At the same time, the sensor nodes have the problems of difficult energy replenishment as well as limited computational resources, and the traditional network structure will be difficult to meet the needs of the next generation grid, so it is of practical significance to study how to improve the energy efficiency of power sensor network. A Mobile Edge Computing(MEC) assisted computing offloading scheme for power sensor network is proposed to optimize the nodes' task processing latency and energy consumption under limited computational resources, by modeling the optimization problem as a Markov Decision Process(MDP) and solving the problem using Double Deep Q Network(DDQN) algorithm to minimize the total system overhead. Simulation results show that the proposed scheme outperforms the benchmark scheme in terms of delay, energy consumption and convergence performance.