Inverse time overcurrent protection is widely used in distribution networks with distributed power sources, but its setting mostly relies on engineering experience, which limits practical engineering application. Thus an optimization method for inverse-time overcurrent protection settings for DG access to distribution networks based on the improved Gray Wolf Algorithm (GWO) is proposed. This is based on the protection action characteristic equation and the requirements for protection sensitivity, speed, and selectivity. In order to overcome the shortcomings of the gray wolf algorithm, such as easy premature maturity and low convergence accuracy, the four types of good point set initialization, convergence factor exponential decay, mutation and heuristic parameter determination are introduced on the premise of retaining fewer adjustable parameters. The simulation example verifies the feasibility and superiority of the model under two-phase and three-phase faults. We discuss the influence of the time level difference on the setting and the direction of future algorithm optimization and engineering application. This provides research ideas for the protection setting optimization problem of large-scale DG access to the distribution network. This work is supported by the Science and Technology Project of China Southern Power Grid Co., Ltd. (No. ZBKJXM20180057).