Aiming at the problem of slow convergence and large errors in the positioning process of the existing wireless sensor network(WSN) optimization algorithm, this paper proposes a Cauchy refraction opposition-based learning and variable helix mechanism of elephant herding localization algorithm. Firstly, the pop ulation is initialized by using Logistic chaotic map with ergodicity and randomness to enrich the population diversity and accelerate the algorithm convergence rate. Secondly, the refraction opposition-based learning mechanism is combined with Cauchy mutation to randomly disturb the position of the patriarch to prevent the algorithm from falling into the local optimum. Finally, an adaptive variable helix strategy is introduced to update the position of ill elephants in the process of clan separation, which improves the global search ability. The simulation results show that the improved elephant herding optimization algorithm proposed in this paper has significantly improved positioning accuracy and convergence rate compared with the existing WSN optimization algorithm.