摘要
气候变化导致西南地区面临愈加严重的干旱问题,为准确预测该区域干旱情势,分别构建了粒子群算法(PSO)和差分进化算法(DE)优化的极限学习机(ELM)模型,并将其应用于云贵高原干旱预测,从预测精度和特征值模拟能力等方面比较了PSO-ELM模型和DE-ELM模型在云贵高原干旱预测中的适用性。结果表明:①PSO-ELM模型和DE-ELM模型均能够有效预测云贵高原不同时间尺度下的干旱状况,模拟值与实测值的相关系数均在0.75以上,均方根误差均小于0.65;②两种模型对最大值、最小值、不同百分位数等特征统计值、干旱历时和干旱强度的拟合效果较好,能够有效反映研究区干旱特征;③相较而言,DE-ELM模型的预测性能略优于PSO-ELM模型,在云贵高原干旱预测中具有更强的适用性。
Climate change has led to a gradual intensification of drought of southwest China.In order to accurately predict the drought regime in this region,the extreme Learning Machine(ELM)models optimized by Particle Swarm Optimization(PSO)and Differential Evolution(DE)were established respectively.In this paper,it used the optimized models(PSO-ELM model,DE-ELM model)to predict drought of Yunnan-Guizhou Plateau.The drought prediction applicability were compared in terms of prediction accuracyand eigenvalue prediction ability.The results showed that①both PSO-ELM model and DE-ELM model could effectively predict drought conditions in the Yunnan-Guizhou Plateau at different time scales,and correlation coefficient between the simulated values and the measured values were above 0.75,and the root mean square errors were less than 0.65.②The two models fit the maximum,minimum,different percentile and drought duration and drought intensity well,which could effectively reflect the drought characteristics of the study area.③The prediction performance of DE-ELM model was slightly better than that of PSO-ELM model,which had stronger applicability in drought prediction of Yunnan-Guizhou Plateau.
作者
赵建华
王华
曹双
ZHAO Jianhua;WANG Hua;CAO Shuang(Suqian Branch of Jiangsu Province Hydrology and Water Resources Investigation Bureau,Suqian 223800,China;Lower Changjiang River Bureau of Hydrological and Water Resources Survey,Bureau of Hydrology,Changjiang Water Resources Commission,Nanjing 210011,China)
出处
《水利水电快报》
2023年第3期15-21,共7页
Express Water Resources & Hydropower Information
作者简介
赵建华,男,高级工程师,主要从事水文水资源分析与论证、防汛防旱等方面的工作。E-mail:zhaojjhh@126.com;通信作者:曹双,男,高级工程师,硕士,主要从事河道治理、防洪研究、水文水资源分析与论证等方面的工作。E-mail:xycaos@qq.com。
