Difficulties programing a model predictive controller using tomlab

Hello, I'm currently programming a non-linear model predictive controller using tomlab's CONOPT as optimizer (third party), the system to be controlled is a polymerization reactor which consists of 8 PDE's, by discretizing in space they turn into 400 ODE's.

My controller is a Simultaneous Approach non-linear model predictive control which means that I use a discretized model of the system as the constraints for my controller. I have two discretized versions of the model, one where I use orthogonal collocation on finite elements to discretize space, and another where I use the WENO finite difference scheme to discretize space as well, in both cases I use orthogonal collocation to discretize time. When I run my controller using CONOPT as solver and using the model where I use orthogonal collocation on finite elements to discretize space, everything runs fine and the optimizer gives satisfactory results, but when I try running the controller with the model discretized using the WENO scheme, it will only give me the same Initial Conditions that I gave to the controller in the first place together with an error message saying "solver error: intermediate infeasible". Do you have any idea what could be causing this behaviour? I know WENO scheme is a far more complex scheme than orthogonal collocation but the optimizer should be able to handle this. In both cases I use the same cost function and I do not provide constraints jacobian nor cost function gradient.