The big issue here is the alignment of your current reference and the DQ axis convention of the PMSM. Your controller is issuing a d-axis current setpoint, and the result is current in both the d-axis and q-axis.
In the convention used for the PMSM model in PLECS, the q-axis is aligned with phase a winding at rotor angle 0 and leads the d-axis by 90 degrees. The magnet flux is on the d-axis, so positive q-axis current produces positive torque. It looks like there a phase shift parameter added to the electrical angle to try and address this, but I don't believe it has the intended effect. Take a look at the DQ current flow in the original model.
Attached is a model where the speed controller output is into the q-axis, and the phase shift parameter you were using was set to zero.
There likely remain some issues for you to sort out, but that was the big change to get positive DC current flow.