Passive source localization utilizing time difference of arrival (TDOA) has widely application in radar, navigation, surveillance, wireless communication, distributed sensor network, etc. This paper presents two robust algorithms named Modified Taylor-seriesmethod (MTS) and Modified Newton (MNT) method. The proposed algorithms are the improvement of the Taylor-series (TS) and Newton (NT) methods for solving the convergent problem which is critical in the iterative methods. The key component of the proposed algorithms is to produce a new modified Hessian matrix intelligently using the Regularization theory which can turn the ill-posed Hessian matrix into a well-conditioned matrix. The regularization parameter which controls the properties of the regularized solution can be automatically determined by the L-curve method. With this procedure, the proposed methods are robust to make the iteration convergence with a bad initial. Simulation results show that the proposed methods improve the convergent probability and have better capability to distinguish the local minimums from the global solutions compared with the TS and NT methods. The proposed methods give superiorperformances of the location accuracy comparing with the closed-form algorithms at large measurement noises.