Abstract: To solve the problem that the multi-source localization in wireless sensor networks is easy to fall into the local optimization as a trap, a new approach to multi-source localization based on Tabu search was proposed the way on acoustic energy model was developed by measuring the signal energy received by several individual sensors, then such parameters as acoustic energy and relevant source location were estimated. To avoid falling into local optimization, the search area including 8 neighbors at 3 levels in grids was introduced, with such parameters as the length of Tabu list and initial estimation selected reasonably. Those sensors with the maximum energy received were the initial estimation nodes. Simulation results showed that in comparison with the multi-resolution (MR) search, EM algorithm and AMR algorithm, the Tabu search algorithm can avoid well falling into local optimization with higher estimating accuracy and lower computation complexity.