This paper introduces a method for inferring temporal uncertain knowledge using Petri nets. First, a temporal knowledge Petri net is created. For every random variable, a corresponding symbolic Petri net is developed. Then, observational places are designed for observational evidence variables, and conditional places are designed to interlink symbolic Petri nets based on their conditional probability relationships. Subsequently, the posterior probability distribution is computed using both the time sequential knowledge Petri net and reachable graph algorithms. Ultimately, by combining this computation with recursion, it is possible to determine the posterior probability distribution for the desired state at time step $ t(t\geqslant1) $, which is achieved by employing time sequential knowledge Petri nets designed to model the dynamic Bayesian network across time steps from $ k $ to $ k+1 $. An example involving battery potential filtering illustrates the proposed method.