Oscillations caused by insufficient system damping may introduce severe instability issues to power systems with a high level of penetration of renewables and electronics. In this paper, repetitive learning control (RLC) is considered for the first time to improve system damping as well as the system dynamic responses. This work introduces an enhanced Automatic generation control (AGC) system with RLC, called RLC-AGC, which helps damp out the oscillations of frequency and tie-line power faster, for both normal load changes and low-frequency inter-area oscillations. The optimal learning gain will be selected based on the derived analytical expressions of the synchronous and damping coefficients, and more importantly, the condition for convergence, to provide a theoretical guarantee for the convergence of the proposed learning control. The New England 39-bus system is simulated to illustrate the performance of RLC-AGC.