Simple code for fitnet-question

The computation of real electrical power between buses 1 and 2 is given by the formula :

Preal= V1^2 * G12 -V1*V2*(G12*cos(delta12)+B12*sin(delta12))

where V1 and V2 are the voltages at buses 1,and 2, G is the conductance and B is the susceptance, delta is the phase angle between buses 1 and 2,

I am training a simple fitnet with just 5 input parameters V1,V2, G12,B12, and delta 12. The target is P real computed by above formula.

After training the net, I want to generalize. This means I am now feeding new input parameters and applying the trained net to these new parameters using the function

newoutput=net(newinput).

I am expecting the newoutput equal to the newcomputed value of Preal using the above formula with newinput parameters.

Instead, I get the newoutput exactly equal to the value of target used using the training phase. Even if I train the net multiple times, I always get the newoutput during generalization equal to the last target value used during the training phase.

What am I doing wrong? Why is the fitnet not generalizing to values of new target values for the newinput parameters?

How can I achieve my desired result?