Abstract:

Traditional on-line optimization ignores the global guidance effect and the search range is not restricted. Hence，such a search has a quality of blindness and can not assure finding the global optimum point，and its solution can not be on-line applied. To overcome these defects，a global-guided flexible tolerance method with two-level optimization algorithm is presented. Genetic algorithm is selected as the global optimization algorithm，flexible tolerance method is for the local optimization algorithm，and the global optimization solution is used to guide the search direction of local optimization. The global optimization and the local optimization each have different periods and constraints. Such a global-guided flexible tolerance method is verified by classical function and applied to the on-line optimization of acetylene hydrogenation reactor. Result showed that this method has the advantages of shorter running time，higher efficiency and more exact solution than the traditional on-line optimization.

摘要：

传统在线优化算法在寻优时未考虑全局最优点对寻优的指导作用，且没有对优化搜索区域进行约束，因此，具有一定的盲目性，很难保证最终能够搜索到全局最优点，且优化解难以在线应用。针对传统优化算法的局限性，本文提出了一种全局优化指导局部优化的两层优化方法——全局校正的可变容差法，以遗传算法作为全局优化算法，可变容差法作为局部优化算法，以全局优化解调整局部优化算法的寻优方向，保证在线优化向全局最优点方向前进，指出全局优化算法和局部优化算法分别具有不同的优化周期和约束区域。将全局校正的可变容差法在数值函数中验证并应用于乙炔加氢反应器的在线优化，结果表明，与传统在线优化算法相比，这种方法不但能够减少寻优时间，也提高了寻优的精确度和有效性。