For the problem that the wireless sensor network partition is still insufficient fault-tolerant after restoring their connectivity, we propose a partitioned dual connectivity recovery method based on Steiner tree and convex polygons. Firstly, the existing leaf nodes are selected based on distance to make a few disconnected discrete nodes unified into zones. The partition is abstracted into points and all non-degenerate quadrilaterals are enumerated. Two Steiner points in the quadrilateral are connected with four vertices to construct Steiner edge deployment relay nodes to make the partition simple connected. Finally, the Graham convex hull algorithm is used to select the convex hull vertices in the abstract points to connect to form convex polygons to realize the dual connectivity of the partition, and the sleeping mechanism is applied to the relay nodes on the second connected path. On the basis of ensuring that the secondary failure of key nodes will not paralyze the network again, the network structure is simplified and the data communication delay is reduced. By simulation, compared with the distributed algorithm for optimized relay node placement using minimum steiner tree(DORMS) and 1C-Sprider Web algorithms, the proposed scheme reduces the number of relay nodes deployed and prolongs the network lifetime.