In this paper, a satellite attitude control with observer-based Modified proportional-integral-derivative (PID) controller is studied in the presence of disturbance and uncertainty. First-order dynamic has been used to model the reaction wheel as control actuator with considering the practical limit of the maximum output torque. In observer method, saturation and windup are feedbacks to control algorithm to modify control signal. Control gains have been obtained by optimization method based on genetic evolutionary algorithm with penalty method and for the performance criterion of the absolute mean of the pointing error. To evaluate the performance, a comparison study has been done between Modified controller and classic controller versus control parameters, phase plane diagram, limit cycle, uncertainties, amplitude and frequency of external disturbance. To fair comparison, all conditions in optimization and initial values are selected identical for two controllers. Comparing results show better performance in the Modified controller, anti-windup, and avoid saturation. In the face of perturbations and limit cycle diagrams, the performance of the Modified controller is clearly comparable to that of a classical controller. In addition, the performance of the two controllers is studied versus moment of inertia, actuator model, disturbance frequency, disturbance amplitude, and maximum momentum uncertainties. The behavior of Modified controller is generally more appropriate and pointing accuracy is better. For example, control accuracy in Modified PID is about 15% better than classical algorithm under moment of inertia uncertainty.