I did a simulation of a dynamic model in Dymola, after the simulation of 200000s is finished, it seems the model hasn't reached a steady state. I am guessing this is caused by the numerical calculation error.
My question is :
How could reach a real steady state when doing dynamic simulation?
Difficult to judge from the single screenshot, but still my thoughts about this:
An exponential decay - which describes many behaviors in a technical systems - will (mathematically) take infinite time to reach its final value. Usually it is assumed, that after five time constants have passed, the process is complete. This corresponds to the value having reached 99.3% of its final value (actually the delta between start and final value). So if you look close enough, you will always find the value changing until you reach so small gradients that they are lost in the numerics.
I do think that the above process is not finished rather than seeing numerical issues here. But again, without a use-case and the rest of the trajectory this is difficult to judge. To shed some more light on this, I would recommend to estimate the time constant of the process (e.g. here). Then it should be possible to judge how "many time-constants have passed" at 2e5 seconds.
BTW: There is a nice website describing the issue for measurements.
