In this paper, a parameter-optimal accelerated iterative learning control (AILC) algorithm is proposed to address the repetitive tracking control for discrete linear time-varying systems within a finite-time interval. In each iteration, based on the learning outcomes from the previous iteration, different extraction matrices are employed to discard data meeting the accuracy criteria, thereby reducing the model's dimensionality and the operational running time, which accelerates the learning control convergence speed and decreases the computational and storage resources required for system operation. The analysis of tracking error convergence leads to the derivation of the control gain design scheme through a parameter optimization method. Finally, simulation comparative experiments between the proposed parameter-optimal AILC algorithm and the traditional iterative learning control (ILC) algorithm verify its effectiveness.