Power system stabilizers (PSSs) are generally used to solve the low-frequency oscillation problem. To overcome this type of oscillation, this paper presented a new method by using the genetic algorithm (GA) to identify PSS settings. The tuning of PSS parameters was formulated relying on an eigenvalue-based objective function aiming at maximizing the stability margin. This was achieved by increasing the sum of the squares of negative real parts of the system eigenvalues. The small disruptions in the form of load change take place routinely; the controller parameters should be adjusted to the changing conditions. To remedy this defect, an adjustment of parameters as a function of the load was required. Thus, a neuronal model by using a historical database determined by the GA solutions for various load levels can approximate the simulation studies, since it could estimate the stabilizer parameters in real time after the learning phase. The validity of the proposed technique was checked through the evolution simulation of the regulator parameters for a load forecast curve. These concepts were discussed in details on a multi-machine network Western System Coordinating Council (WSCC) comprising three generators and nine nodes. The eigenvalue analysis and time domain simulation results were presented at different operating conditions and under various disturbances in order to show the effectiveness of this study.