The feedback maximization of reliability of simple power systems under random perturbations was studied.The stochastic averaging method for quasi non-integrable Hamiltonian systems and stochastic dynamical programming principle were applied to derive the dynamical programming equation for the maximization of conditional reliability function and that for the maximization of mean first passage time.The optimal control, the conditional reliability function and mean first passage time of uncontrolled and controlled power system were obtained by solving the corresponding final dynamical programming equations.The proposed method was verified by using Monte Carlo simulation.