A non-contact acoustic pressure-based method is proposed for load identification in the acoustic structure interaction system involving non-probabilistic uncertainty. The forward problem for load identification is established through the discretized convolution integral relationship of the dynamic loads and the Green’s kernel function matrix of the system. The inverse process is constructed by using truncated single value decomposition approach in order to overcome the ill-posedness of the global kernel function matrix. In this work, two non-probabilistic models including ellipsoid model and interval model are proposed to quantify the effects of the system uncertainty. Several numerical examples are investigated to verify the effectiveness of the present methods. The results show that the non-contact acoustic pressure-based method with great convenience for dynamic load identification is accurate and effective. For the system with non-probabilistic uncertainty, the ellipsoid model and interval model are the proper choices to identify the bounds with knowing only the extreme values of the parameters. Moreover, the load bounds derived from the ellipsoid model are more reliable than those derived from the interval model.