The novel power system adopts the Internet of things architecture to deploy the new architecture of the power system, which is the key technology for the development of traditional power grid to transform to new energy Internet. How to make real-time, efficient and low-cost data transmission for the massive data generated in power grid is a key issue. In order to solve the contradiction between transmission cost and transmission rate in the process of power grid data transmission, a low-cost and efficient power data transmission method based on compressed sampling matching pursuit (CoSaMP) technology is studied in this paper. The original data of power grid is randomly sampled and compressed. The compressed data is transmitted. Finally, the compressed data is restored by reconstruction algorithm. The proposed method is verified by a simulation example, and the proposed method is compared with several traditional data reconstruction algorithms. The simulation results show that the proposed CoSaMP method can restore the original signals and effectively improve the effectiveness of power grid data transmission within the allowable errors.