We propose a model-based state-of-charge (SOC) estimation method for rechargeable batteries which guarantees finite-time estimation error convergence. One of the most common approaches for the SOC estimation based on equivalent circuit model is to utilize the Kalman filter. The method has an asymptotic convergence property. From the practical standpoint, however, an estimation method that guarantees a finite-time estimation error convergence is desired. To this end, we employ the impulsive observer which achieves the finite-time estimation error convergence with the exact model. Furthermore, we analize the bounds of estimation errors in the case that the model used in the impulsive observer has errors. Finally, the effectiveness of the proposed method is illustrated through numerical experiments.