Aiming at the observed data of step-stress accelerated degradation test (SSADT) in similar products, a nonlinear Wiener process is used to build an accelerated degradation model with measurement error. Two-step maximum likelihood estimation method is used to obtain the fixed coefficient estimation and random coefficient prior in the degradation model. Based on the probability distribution of the first hitting time (FHT), the approximate expression of the probability density function (PDF) of the remaining life is deduced. By using the current observed data of the target product, the Bayesian updating method with random coefficients is introduced to update the posteriori value of the random coefficient so that the remaining life prediction can be updated online. Combined with a certain laser simulation example, the correctness and advantage of the proposed model are verified.