基于极限学习机的干旱区潜在蒸发量模拟

Numerical Simulation of Evaporation of Arid Region Based on Extreme Learning Machine

准确地模拟干旱区潜在蒸发量,对区域水资源的合理开发利用与生态环境保护具有十分重要的意义。以极限学习机(ELM)模型为基础,以古浪河流域的乌鞘岭、古浪两个典型气象观测站点为对象,将气象因子的不同组合作为输入参数,构建了适合当地的潜在蒸发量模型。利用构建的模型对乌鞘岭、古浪气象观测站点的月潜在蒸发量进行了模拟,将模拟结果与支持向量机(SVM)模型模拟结果进行了对比,发现ELM模型在干旱区月潜在蒸发量模拟中有更好的适用性,可为干旱地区潜在蒸发量的估算提供新方法和思路,是资料有限条件下潜在蒸发估算的有效方法。

The simulated accurately evaporation at the arid regions was essential to rationally develop and utilize water resources and the ecosystem protection.The meteorological data at Wushaoling and Gulang weather stations in Shiyang River Basin were used in this study.Based on the extreme learning machine（ELM）model,the varying combination of meteorological factors were inputted to the model.An evaporation model also was established to simulate monthly evaporation at the two weather stations,and the results were compared with support vector machine（SVM）model to evaluate the simulation ability of ELM model.Our study demonstrated that the ELM model had better applicability in simulating monthly evaporation at arid regions.It can provide a new method and idea for calculating evaporation,and it is a desirable and effective method to calculate evaporation at the arid regions with insufficient data.