LQR Reference Signal for HSV Reduced-Order Model

I am attempting to implement an LQG controller using a reduced-order state-space model. The kalman filter has already been implemented, and the controller works well; however, it is only able to track the operating point around which the original system was linearized. In short, here is what I have done:

1) Linearized non-linear equations of motion around a fixed operating point x_op

2) Applied a model reduction scheme via matlab model reduction to reduce number of states from 107 to 13.

3) Implemented a Kalman filter on the reduced model.

4) Designed an LQR controller for the reduced model.

I know the reference I want to track in the original model (x(10) = ref), but I don't know what to set the reference to for the reduced model. Since the actual state values are no longer what they were for the full model, and I am not sure how to determine what it should be. Additionally, I suspect that due to coupling of states, it will be a projection of the reference signal onto the new state vector. How can I determine the new reference target?