A Multi-objective Optimization Strategy Based On Kriging Surrogate Model And Its Application In PX Oxidation Process

A multi-objective optimization problem does not exist the optimal solution which is similar to the single-objective optimization problem, but a set of non-inferior solutions. Thus extending the signal-objective particle swarm algorithm to the field of multi-objective optimization, determining the global extremum becomes a difficulty. In addition, the basic multi-objective particle swarm algorithm always has bad convergence and distribution, which is also easy to fall into the local optimal solution. This is because of the diversity of the population can’t be maintained. Therefore, to improve the diversity of the population is also very important.For the problems of multi-objective particle swarm optimization encountered, a novel particle swarm optimization used in solving multi-objective optimization problem is proposed in this paper. This proposed algorithm adopts a novel strategy to select the global best particle and the individual best particle, in order to improve the stability of moving forward to the true Pareto Front. At the same time, this algorithm also uses an external archive to store non-dominated solutions at each generation and uses the dynamic update crowding distance to keep the size of the external archive. Additionally, for the sake of improving the ability to escape from local optimal solutions, this proposed approach also introduces a novel mutation with two steps. And Then, conduct experiments on a set of classical ZDT multi-objective test functions and compare with MOEA/D、NNIA and NSGA-Ⅱ. The results show that this algorithm compared with other algorithms has better convergence and distribution.At the end, this algorithm is used in PX oxidation process which compares with other multi-objective algorithms and reduces the Acetic acid and PX combustion loss obviously under the same cost. At the same time, we find that based on the flow path simulation model, complex chemical process optimization often takes a long time to optimize and has low efficiency. This paper puts forward using Latin hypercube sampling and Kriging modeling method to build a surrogate model of the flow path simulation model. Then a multi-objective optimization strategy based on Kriging surrogate model is established. Using this strategy in PX oxidation reaction process optimization, Optimization time is decreased greatly, and the efficiency of optimization is improved obviously.