蝙蝠算法优化极限学习机模拟参考作物蒸散量

Simulation of reference crop evapotranspiration by using bat algorithm optimization based extreme learning machine

为提高参考作物蒸散量模拟的准确性,提出蝙蝠算法优化极限学习机的参考作物蒸散量模拟模型.基于汕头站1966—2015年月值气象数据(包括逐月最高温度、最低温度、地表总辐射量、风速和相对湿度),建立参考作物蒸散量的极限学习机模型,并采用蝙蝠算法通过交叉验证方法对极限学习机的正则化系数和径向基函数的幅宽进行优化,最后对参考作物蒸散量模拟效果进行评估.结果表明:与传统调参方法和遗传算法优化后的模型相比,蝙蝠算法优化参数极限学习机模型建立了整体性能优异并且稳定的参考作物蒸散量模型,提高了参考作物蒸散量的模拟精度.

In order to improve the prediction accuracy of modelling reference crop evapotranspiration, the bat algorithm was used to optimize extreme learning machine （ELM）. Meteorological data from 1966 to 2015 at Shantou Station （i. e. , monthly maximum and minimum ambient temperatures, global solar radiation, wind speed and relative humidity） was used to train and test the proposed models of extreme learning machine. The bat algorithm was used to optimize the regularization coefficient and breadth of radial basis function of ELM with a cross-verification method. Finally, the performance of proposed models for the reference crop evapotranspiration estimation was evaluated by statistical indicators. The results show that the bat algorithm - based optimized ELM model provides better accurate and stable values of reference crop evapotranspiration in comparison with the evapotranspiration values estimated by the models optimized with traditional tuning method genetic algorithm, respectively.