基于代理模型的常压塔过程多目标优化

Surrogate Based Multi-objective Optimization for Atmospheric Tower

针对常压塔过程进化优化时存在的耗时计算问题,提出采用RBF、Kriging代理模型研究常压塔操作过程的多目标进化优化。在优化迭代中分别采用了"最优点"及"期望值提高"(ExI)两种插值方法,并采用了模糊规则对代理模型进行了调用。优化结果表明:采用代理模型可以减少进化中对机理模型的调用次数,从而提高优化效率;对常压塔来说,与RBF的"最优点"插值方案相比,采用Kriging的ExI点插值建立的代理模型精度更高,优化效果也更理想;调节常压塔的原油混炼比及其它操作条件可以在相同的CO2排放下提高经济效益。

Aiming at the time-consuming computation problem of process optimization in atmospheric tower,in this paper,a multi-objective evolutionary optimization strategy of atmospheric tower operation process is studied based on RBF and Kriging surrogate models respectively.In the optimization iteration,two interpolation methods,namely the’best point’and’expected improvement’are adopted respectively,and the fuzzy rules are used to manage the surrogate models.The optimization results show that using surrogate model can reduce the number of invocation to the mechanism model in evolution,and thus improve the optimization efficiency.For atmospheric tower,the surrogate model established by Kriging’s ExI interpolation has higher precision and better optimization effect compared with the RBF’s’best point’interpolation strategy.The adjustment of the crude oil mixing ratio and other operating conditions of the atmospheric tower can improve the economic efficiency under the same CO2 emission.