建立基于模拟退火遗传算法(Simulated Annealing Genetic Algorithm,SAGA)的改进极大似然法,即将似然函数相反数求解极小值的表达式作为目标函数,依据矩法估计参数取值范围作为约束条件,然后应用SAGA进行参数估计。与常规极大似然法思路有本质不同,改进极大似然法通过遗传算法进行参数优化。通过蒙特卡罗试验,验证了改进极大似然法在参数估计和不同频率设计值估计两个方面均具有很好的准确性,与基于最大熵原理的方法效果相当,优于其他方法;同时该方法不受线型类型、参数数目和约束条件的限制;可以避免应用常规极大似然法时出现似然方程无解等情况;且求解过程简便快捷,使极大似然法在理论上和实际应用中都成为有效的方法。
In this paper, an improved method for parameters estimation in hydrologic frequency analysis, SAGA-ML, was proposed. The idea of SAGA-ML is establishing one parameter optimization program whose object function is the minimal likelihood function of hydrologic probability distribution, and the constraint equations are gotten by MOM, then SAGA (Simulated Annealing Genetic Algorithm) is used to optimize parameters. The SAGA-ML method can overcome the disadvantages of normal maximum likelihood method because of using SAGA for parameter optimization. By Monte-Carlo tests, it has been concluded that the SAGA-ML method is effective and convenient when being used. The SAGA-ML method is suitable for any hydrologic probability distributions, any numbers of parameters and any constraint conditions, so it is useful and effective both in theory and practice.