As the scale of power grid business continues to expand, smart grids are rapidly emerging as an alternative to traditional power grids. In order to realize the efficient processing of power grid data, cloud computing and edge computing have been widely used in smart grid. However, how to realize the unified arrangement between edge nodes and cloud nodes also needs to be solved. To this end, Compute First Networking (CFN) was proposed, whose goal is to achieve optimal and efficient use of network and computing resources. However, how to efficiently obtain computing and network resource status in the smart grid is a major challenge, due to equipment damage, disconnection or even attack. Therefore, this work focuses on reconstructing complete computing and network resource information under incomplete status in smart grids and propose a resource sensing method based on deep matrix completion which can effectively obtain computing and network resource information. The experimental results evaluate the accuracy of our proposed method.