The accuracy of load identification is often hindered by the inversion of an ill-conditioned transfer function matrixs at frequencies near the structural resonances. To overcome this inversion instability, total least squares (TLS) method is a successful approach for linear problems is introduced. Tikhonov regularization of the TLS leads to an optimization problem of minimizing the sum of fractional quadratic and quadratic functions. Then, a conjunction gradient(CG) method is proposed for solving the tikhonov TLS optimization problem, called CG-TLS algorithm, whose advantages are more simple to implement，smaller storage capacity, better convergence performance and that can consider not only vibration response but the transfer matrix is contaminated by noise. The ill-conditioned reasons of transfer function matrix are investigated by numerical simulations and test of load identification, then choosing the locations of vibration response optimally with condition number. Finally, The CG-TLS regularization algorithm and other two methods are used to identify vibration load in different noise levels. The results demonstrate that the CG-TLS regularization algorithm outperforms general regularization methods showed the best performance, which also has lower noise sensitivity. Therefore the new algorithms established in this paper has broad prospect of engineering application.