Hi Will,
Thanks for your kind reply. I have to admit that the case is actually a 3d problem and the loss of accuracy is inevitable. In fact, one reason I set the case in 2D is that I’d like to check my mesh was ok for the problem, so then I could extrude it in 3D with the least calculation cost. The other is I could compare with the result to weight how much the three-dimensional effect can be. So I will try a 3d simulation and see if any improvement can be achieved.
As for the regular pattern of the time history curve, I would be expecting a periodic sine-like curve rather than a random one. The paper you attached is quite impressive and has a certain similarity to what I’ve got. 
But from the time history curve of Cl, It’s a bit strange to see the average of Cl is non-zero, which may happen in the initial unstable period I think. I’ve looked through a few theses, and found the result showed a periodic curve at different degree, though most of them were 3d cases. Just weeks ago, I tried running circular cylinder cases with a relatively lower Reynolds number, and the results were pretty good. And that’s why I was surprised when not find the regular shedding of large eddies.
Last but not least, I did gain a lot from the p-multigrid and Niki’s replies from other posts. But It seemed that the pressure residual was choked and cannot reduce to below 0.1 even for a long period. Or maybe the time is not long enough? On the contrary, from the reply in the other post, with 3 cycles, the level of convergence of the turbulent Jet can be below 1e-3. I will try with longer time and a higher ac-zeta(if it’s necessary).
P.S. I found that sometimes the anti-alias = (flux, surf-flux) option can raise a Keyerror : Kernel mul has no providers, I would open a new topic if there exist any problem. Sorry for my being tediously long, I just want to explain everything I think is relevant. 
Regards, Thatcher.