改进的实数编码加速遗传算法在暴雨强度公式参数优化中的应用

针对城市暴雨强度公式参数识别时,传统求解方法(如牛顿迭代法、高斯-牛顿法等)存在间接拟合,而优化算法(如实数编码加速遗传算法、蚁群算法等)存在随机性,盲目地在一个区间内寻优拟合精度不高等问题,本文将两种方法结合,提出一种改进的实数编码加速遗传算法(RAGA),为暴雨强度公式参数识别提供一种新途径。该方法将传统求解方法所求的可行解作为改进遗传算法的初始参数,通过在每次代际寻优时设置各参数廊道约束来改进RAGA以提高算法搜索效率,直至公式拟合精度无法提高为止。将该方法应用于国内多地暴雨强度公式参数识别中以评估算法的有效性,结果表明此方法实用可行、搜索效率较高,可以快速收敛到最优解。实例表明该方法在暴雨强度公式参数识别中是实用有效的。

Improved real coding-based accelerated genetic algorithm and its application in the parameters optimization of storm intensity-duration-frequency relationship

There are commonly two methods used in the parameter optimization of storm intensity-duration-frequency ( IDF) relationship，conventional numerical approach and traditional optimization algorithm． The conventional numerical approach，such as the Newton iteration and Gauss-Newton algorithm，usually suffers from a relatively low accuracy owing to its indirect fitting of the parameters，while the traditional optimization algorithm，for example，the real coding-based accelerated genetic algorithm ( ＲAGA) and ant colony algorithm，often undergoes a low efficiency due to its randomness and blind searching in a certain interval． Based on the above two methods，we propose a new improved ＲAGA for the parameter optimization of storm IDF relationship， which sets the feasible results of the conventional numerical approach as the initial parameters，and can improve the precision of parameters evaluation by setting the corridor constraints in each optimal iteration． Applications in three different meteorological stations of China have demonstrated that the new approach is more practical and efficient in the parameter optimization．