Energy consumption of crude oil distillation unit accounts for about 25% to 30% of refinery total energy. Satisfying the request of product yield and quality

optimizing crude distillation column operating conditions can effectively reduce energy consumption. Using stochastic optimization algorithms directly optimizing model of atmospheric tower is time-consuming and low efficient. In this paper

efficient global optimization algorithm based on surrogate model is applied to optimization of atmospheric tower's heat recovery. Kriging surrogate model is used as a replacement to the original model which is time-consuming in iterative optimization process. The result shows that this method decreases 90% number of assessment and decreases 85% optimization time compared with particle swarm optimization and realizes energy savings and meets product separation accuracy requirements.