关键词: 过程系统优化, 元模型, 均匀设计, 支持向量机, 遗传算法

Abstract: A general methodology for optimization of complex process system via empirical meta-modeling is described. The uniform design and support vector machine are used to build a meta-model of the system, and the meta-model as fitness function is incorporated into genetic algorithm. This optimization methodology involves data collection from the process simulation model or real operation, and fitting to less complex surrogates: meta-model, which is more readily optimized. The use of empirical meta-model allows the optimization to complex process while requiring only a few of solutions to be obtained from the process model. The effectiveness of proposed optimization methodology of complex process system has been proved by optimizing the operating parameters of Puguang high acid natural gas purification plant. The meta-model of whole process has been built by selecting 10 experimental points within the optimizing space of operating parameters. The relative error of predicted value by the meta-model is less than 4%. The optimization results show that the exergy efficiency of purification plant could be increased by 6.6% under the optimum operating point.

Key words: process system optimization, meta-model, uniform design, support vector machine, genetic algorithm