If you want to change your boundary condition during a transient simulation, you can assign a time dependent boundary condition. For that, you can write a user defined function (UDF) and hook it at appropriate place in the "boundary condition" panel.
According to me and by definition, boundary conditions do not change during a step (or steps) of for example a time series. That's why they are boundary conditions in the first place. If they change you assume that the bio-geophysics of the processes determining your time series, change their nature (definition) as well. I guess in that case you need another approach (mathematical model) in the new case. If a Fourier series fits a certain proces time series well, its (mathematical) boundary conditions stay valid for the complete time series. If you see a bad fit from a certain timestep. I think, you need another temporel series model (polynomial, probabilistic math time series models etcetera...).
If you want to change your boundary condition during a transient simulation, you can assign a time dependent boundary condition. For that, you can write a user defined function (UDF) and hook it at appropriate place in the "boundary condition" panel.