Abstract: The load time variability constrains the development of the load modeling, but ambient signal based load modeling provides a new solution to overcome this problem. However, the load random fluctuation cannot be ignored in the ambient signal environment and leads to the difference between measured data and the typical load model, which also affects the results of ambient signal-based load identification. Therefore, the impacts of internal disturbances in different frequency bands on load response properties under ambient signal excitation are analyzed in this paper to explain the reasons of differences between measured data and the typical load model. Then, an identification method considered filtering is proposed. Finally, the load properties analysis and improved method are verified based on simulation, the results indicate that the filtering can extract effective load responses and improve effect of ambient signal-based load modeling.