基于并行自适应遗传算法的水文模型率定研究

Research on hydrologic model calibration based on parallel adaptive genetic algorithm

【目的】参数率定是影响水文模型预报精度的重要因素,采用人工智能算法可以有效提高水文模型参数的率定效果。【方法】采用基于种群离散程度的自适应算子,对GA算法的交叉、变异和迁移过程进行自适应优化,并利用粗粒度并行计算模型提高种群进化效率,综合以上手段研究了一种基于自适应策略的并行遗传算法。将传统遗传算法(GA),串行自适应遗传算法(AGA)和并行自适应遗传算法(PAGA),应用于屯溪流域新安江模型的参数率定,从率定效率、率定收敛性、率定稳定性和率定效果四个方面,验证PAGA算法的综合性能。【结果】结果表明:PAGA算法的计算加速效果显著,在10核环境下相对于AGA算法计算时间减少了87.9%;在进化后期,PAGA算法能够更加稳定的收敛于最优解,收敛后的目标函数值具有更好的稳定性;在验证期的场次洪水模拟中,采用PAGA算法率定的模型模拟效果最优,总体洪水合格率大于90%,确定性系数均值为0.85。【结论】PAGA算法能够明显降低模型参数寻优耗时,改善模型率定效果和收敛性能,为水文模型参数的率定提供了新思路。

[Objective]Parameter calibration is an important factor affecting the accuracy of hydrological prediction.The intelligent algorithm can effectively improve the calibration effect of hydrological model parameters.[Methods]An adaptive operator based on population dispersion is adopted to optimize the crossover,variation and migration process of genetic algorithm.The coarse-grained parallel computing model is used to improve the efficiency of population evolution.Based on the above method,a parallel genetic algorithm based on adaptive strategy is proposed.The traditional genetic algorithm(GA),serial adaptive genetic algorithm(AGA)and parallel adaptive genetic algorithm(PAGA)were respectively applied to the parameter calibration of Xin′anjiang model in Tunxi Basin.The comprehensive performance of PAGA algorithm is verified from four aspects of calibration efficiency,calibration convergence,calibration stability and calibration effect.[Results]The results show that the PAGA algorithm has a remarkable acceleration effect,and the calculation time is reduced by 87.9%compared with AGA algorithm in the 10-core environment.In the later stage of evolution,PAGA algorithm can converge more stably to the optimal solution,and the value of the objective function after convergence has better stability.In the verification period,the model optimized by PAGA algorithm has the best simulation effect,the overall pass rate of flood simulation is greater than 90%,and the average certainty coefficient is 0.84.[Conclusion]PAGA algorithm can obviously reduce the time of parameter optimization,improve the model calibration effect and convergence performance,and provide a new idea for the hydrological model parameter calibration.