Aiming at the problem that most optimized designs for accelerated degradation tests do not consider the impact of cognitive uncertainty introduced by different sample sizes on test results, a new method for optimal design of accelerated degradation tests under Weibull distribution is proposed. A deviation index is constructed to characterize the cognitive uncertainty caused by different sample sizes. Based on this, an accelerated degradation test optimization design model with the Minimum Relative Error (MRE) as the goal is established. The model takes the total sample size, the number of accelerated stress levels and the proportion of sample distribution as design variables. The solution algorithm of the optimization model is given, and the proposed method is applied to the optimized design of accelerated degradation test of a certain type of laser, which shows the effectiveness of the method.