Microstrip reflectarray antennas (RAs) are able to provide equivalent performance of a traditional parabolic reflector, but with simple and light electromagnetic and mechanical structures; this can be achieved by an effective control of the phase response of each microstrip element on the reflecting surface. Thus the main problem is the fast and accurate modelling of the unit microstrip patch to be used in the array optimization. Since using simulators in the optimization procedure is computationally very ineffective, herein artificial intelligence is used as a rapid and accurate tool for characterization of the reflection angle of the microstrip unit element in terms of its geometry, substrate parameters and frequency. For this purpose, modelling of Omega shaped patch within the X-band is considered using Multilayer Perceptron (MLP) neural network trained the 3D CST microwave Studio simulator data. Validation of the MLP model is also worked out successfully with the 3D CST data. Thus a continuous function is obtained for the reflection angle in the geometry, substrate and frequency domain of the microstrip patch unit element that ensures the reflectarray optimization procedure as fast as using the analytical functions and as accurate as the 3D simulators. In the paper, full-wave simulation- based MLP modelling of the omega shaped microstrip patch is given in details.