The invention relates to the wireless sensor network technology field and discloses a fault recovery method of a wireless sensor network. In the invention, through a leading relay node (LRN), an isolated area which loses connection is reconnected to a nearest connected set (CS), and then the CSs are connected so as to realize a purpose of recovering network connectivity. Because the LRN in the each isolated area only needs to move to the CS which is nearest to the LRN itself, a shift distance can be substantially shortened and node energy can be saved too. In the invention, according to a density of the LRN, CS division is performed to the whole network and simultaneously, during a connection process of the different CSs, a density approach method is still used to perform the fan-shaped division so that an LRN optimization process between the isolated areas can be simplified and complexity of algorithm can be reduced.