Production planning and inventory management as an important component of production management, play an important role in production and operation of enterprises. Based on the oil demand forecast and taking the economic benefits of enterprise-scale into account, multi-stage production inventory optimization model which is based on the changes of processing costs is established to minimize the production costs. The model demonstrate not only the impact of unit processing cost on the production plan, but also consider the finished goods inventory optimization, therefore, meet the business requirements of minimizing inventory costs. The adaptive weight particle swarm algorithm and simulated annealing-based particle swarm algorithm are used to solve the model, The experimental results show that simulated annealing based on the performance of PSO is better than the adaptive weight of the particle swarm algorithm, which can jump out of the local optimum and converge to the global optimum faster.