Try to optimize by using Z -matrix format of the given molecule.

In theory the Symmetry(PG=D2d) command should do the trick, though I never used it. 

In general, after the D4h optimization finishes you can manually distort the geometry to D2d and rerun it - it should recognize the new symmetry automatically.

Hi,

Check these links, I hope you will find the answer.

http://www.gaussian.com/g_tech/g_ur/k_symmetry.htm

https://www.researchgate.net/post/How_can_one_perform_a_symmetry_constrained_optimization_in_gaussian

I agree with Bartosz Trzaskowski. After the D4h optimization , distort the molecule by changing the suitable bond length or angle and then optimize. 

Or View the molecule in the Gauss view and set the point group.

Or Put symm=loose in the keyword section and optimize

Hope this will work

If you are working on systems with Jahn-Teller distortion, the best way is to specify the coordinates in a Z-matrix that has the symmetry you want. If you rely on using keywords, you may run into troubles during geometry optimizations, should the detected symmetry change significantly. 

# opt freq b3lyp/6-31+g(d,p) scf=xqc symmetry=(follow,pg=D2h)

I have used this for anthracene optimisation in z-matrix format. The output came in D2h format as I was expecting. I have tried with “iop” and various other ones, but none gave the output with symmetry other than C1or C2v.