The saturated PI (proportional-integral) based method is widely applied in nonlinear system control fields. It can be regarded as a black-box type approach which utilizes the system output's tracking error and its integral information, with the saturated control input. However, as details of the plants may not be necessary to investigate, the saturated PI control methods has to empirically tune the proportional and integral parameters to guarantee reliable convergence, making its convergence mechanism can not be generally interpreted. In this brief, for the first time, the convergence of the saturated PI control scheme is proved through the optimization solver based on a primal dual neural network. Illustrate examples including control of an inverted-pendulum mobile vehicle and a manipulator demonstrate the efficiency of the saturated PI control methods based on the proposed optimization formulation.