In order to optimize the localization algorithm and improve the accuracy of positioning in Wireless Sensor Network(WSN), a kind of weighted centroid localization algorithm based on multilateral localization error was proposed. After analyzing the transmission model of wireless signal, a model of distance measurement was established on account of logarithm fit about the relationship between Received Signal Strength Indicator(RSSI) and distance. Then the multilateral localization algorithm and location estimation model of solving unknown node′s coordinates were introduced. After the orientation of several sets of data, the reciprocal of positioning error was used as a weight in the process of calculation to improve usual centroid algorithm and the influence of selection of reference point number to the error was discussed. The experimental result shows that when compared with traditional centroid algorithm, the improved weighted centroid algorithm has better localization precision, when choosing four or five reference nodes, the experiment can achieve better location performance.