Hi Chi Chun,
The dimensional analysis ensures that the units in your model are consistent. However, this requires you to specify units for every compenent of your model (species, compartments, parameters), and for them to be consistent. See the screenshot (attached), you are missing the unit for the cells and the unit for kon is inconsistent. You can use the reaction rate to determine which units you should use. E.g. your reaction is P+L <-> PL. The reaction rate is P*L*kon-PL*koff. The unit of a reaction should be in amount or concentration per unit time. So each term in the reaction rate should have those units. Given that you defined P, L and PL in micromole (amount), the unit for kon needs to be 1/micromole/picomole. Once you have added/changed the units, you can verify the model to make sure that the unit conversion works and the units are consistent.
On another note, it looks like one of your reactants (Streptavidin) is quadrivalent. Looking at the reactions defined in the model, there are different reaction rates depending on how many binding sites are occupied. To model that, I suggest having species for P, L, PL1, PL2, PL3, PL4. So far you have implemented the reaction between P, L and PL1. From there you can create a network of reactions to model the complete binding of P to L, as described in the model. You'll also need to define koff and kon rates for each reaction.