3D Poisson equation problem coupled with Nernst-Planck equation

I used coupled Electrostatics physics and Transport of Diluted Species physics to simulate this problem.
The source term in Poisson equation is the difference between cations and anions.
Usually, the difference should be zero due to the electroneutrality. In some conditions, it would be not.

My problem is that, when I used 2D model, the simulation converged even with a coarse mesh. When it comes to 3D situation, it no longer converged, no matter how fine the mesh is.(as far as I think, the mesh is fine enough)

The error is as below:

Failed to find consistent initial values.
Segregated group 1

Undefined value found.
Undefined value found in the stiffness matrix..
For mesh-case 1 there are 111061 equations giving NaN/Inf in the matrix rows for the variable mod1.H.
and similarly for the degrees of freedom, NaN/Inf in the matrix columns.
Last time step is not converged.

I'm not clear about the numerical methods. It seems that the solving processes of 2D and 3D are quite different. 3D used segregated groups, which is not shown in 2D process.

I need your help.
Thanks in advance.
Mingjie Jia