Abstract We investigate an adaptive neuro-control approach, namely goal representation heuristic dynamic programming (GrHDP), and study the nonlinear optimal control on the multi-machine power system. Compared with the conventional control approaches, the proposed controller conducts the adaptive learning control and assumes unknown of the power system mathematic model. Besides, the proposed design can provide an adaptive reward signal that guides the power system dynamic performance over time. In this paper, we integrate the novel neuro-controller into the multi-machine power system and provide adaptive supplementary control signals. For fair comparative studies, we include the control performance with the conventional heuristic dynamic programming (HDP) approach under the same conditions. The damping performances with and without the conventional power system stabilizer (PSS) are also presented for comparison. Simulation results verify that the investigated neuro-controller can achieve improved performance in terms of the transient stability and robustness under different fault conditions.