Abstract:

Acetylene hydrogenation reactor is an important device of the ethylene plant, and its working condition directly affects the purity and yield of ethylene products, so it is of great significance to improve the modeling and optimization efficiency of the reaction process for acetylene hydrogenation. However, directly optimizing high precision CFD models are often low efficiency because of the large calculation. This paper from the model accuracy of prediction variance and improvement of the effectiveness of global search, proposes a bootstrap GEI algorithm based on Kriging surrogate model. Through simulation contrast of test functions, compared with bootstrap EI algorithm, this algorithm can find the optimal sample points of global search and reduce the number of sample points, so as to improve the efficiency of the model updating and optimization. And this algorithm is effectively validated in the surrogate model of acetylene hydrogenation reactor.

Key words: model, EGO, bootstrap algorithm, hydrogenation, simulation

摘要：

碳二加氢反应器是乙烯工厂的一个重要装置,其运行状况直接影响着乙烯产品的纯度和产量,因此,提高碳二加氢反应过程的建模和优化效率对于工业实际具有重要意义。但是,直接对碳二加氢反应器的CFD高精度分析模型进行优化,计算量大,优化效率非常低。本文从模型预测方差准确性和提高全局搜索的有效性出发,基于Kriging代理模型,提出了一种求取无偏预测方差的广义期望提高算法——bootstrap GEI算法,通过测试函数的仿真对比,与bootstrap EI算法相比,该算法能够从全局角度搜索最优样本点,减少样本点的个数,从而提高模型更新和优化效率。该算法在实际工业碳二加氢等温反应器的代理模型中也得到了有效验证。

关键词: 模型, 全局优化算法, bootstrap算法, 加氢, 模拟