APP下载

基于粒子群优化(PSO)超限学习机预测新疆参考作物蒸散量

2021-07-23尹起周建平许燕李志磊樊湘鹏魏禹同

江苏农业学报 2021年3期
关键词:新疆

尹起 周建平 许燕 李志磊 樊湘鹏 魏禹同

摘要: 參考作物蒸散量(ET0)的准确预测对于作物需水量预测、农田精准灌溉和提高水资源利用效率等具有重要意义。为了解决传统方法获取ET0的弊端,本研究基于粒子群优化(Particle swarm optimization,PSO)-超限学习机(Extreme learning machine,ELM)预测ET0。通过选取新疆地区3个站点(乌鲁木齐、喀什、哈密)的最高气温(Tmax)、最低气温(Tmin)、平均相对湿度(RH)、风速(u2)、光照时间(n)等气象数据,建立PSO-ELM预测模型,对模型精度和普适性进行研究,并通过与ELM、Makkink、I-A模型的对比,探究不同气象因子组合模型的预测精度。结果表明,PSO-ELM模型在5种气象因子输入下具有最高预测精度(平均R2=0.974 7,平均MAE=0.252 0 mm/d,平均RMSE=0.364 3 mm/d)。由PSO-ELM6模型与ELM、Makkink、I-A模型的对比结果看出,在相同的气象因子输入条件下,3个站点用PSO-ELM6模型预测的效果最好。通过对PSO-ELM3模型在新疆地区普适性的研究发现,该模型具有较高的预测精度(平均R2=0.946 5,平均MAE=0.307 0 mm/d,平均RMSE=0.356 9 mm/d)。不同站点、不同气象因子输入的PSO-ELM模型能够较为精准地反映气象因子与ET0之间复杂的非线性关系,且模型在新疆地区的普适性较好,可以为新疆地区逐日ET0预测提供新的方法。

关键词: 新疆;粒子群优化;超限学习机;参考作物蒸散量;模型精度

中图分类号: S27;TP312 文献标识码: A 文章编号: 1000-4440(2021)03-0622-10

Prediction of reference crop evapotranspiration in Xinjiang based on particle swarm optimization(PSO) optimized extreme learning machine

YIN Qi1, ZHOU Jian-ping1, XU Yan1, LI Zhi-lei2, FAN Xiang-peng1, WEI Yu-tong1

(1.College of Mechanical Engineering,Xinjiang University,Urumqi 830000,China;2.Engineering Training Center,Xinjiang University,Urumqi 830000,China)

Abstract: Accurate prediction of reference crop evapotranspiration (ET0) is of great significance in predicting crop water demand, precise irrigation of farmland and improving water resource utilization efficiency. To solve the disadvantages of traditional methods in obtaining ET0, ET0 was predicted based on particle swarm optimization (PSO)-extreme learning machine (ELM) in this study. By selecting meteorological data such as maximum temperature (Tmax), minimum temperature (Tmin), average relative humidity (RH), wind speed (u2) and illumination time (n) of three stations in Xinjiang (Urumqi, Kashgar and Hami), the PSO-ELM prediction model was established. The accuracy and universality of the model was studied, and the prediction accuracy of models combined with different meteorological factors was explored by comparing with ELM, Makkink and I-A models. The results showed that, PSO-ELM model showed the highest prediction accuracy under the input condition of five meteorological factors (average R2=0.974 7, average mean absolute error=0.252 0 mm/d, average root mean square error=0.364 3 mm/d). The prediction effect of PSO-ELM6 model was the best under the same meteorological factor input conditions of three stations by comparing the PSO-ELM6 model with ELM, Makkink, I-A models. The research on the universality of PSO-ELM3 model in Xinjiang showed that, the model had high prediction accuracy (average R2=0.946 5, average mean absolute error=0.307 0 mm/d, average root mean square error=0.356 9 mm/d). The PSO-ELM model with different meteorological inputs at different stations can accurately reflect the complex non-linear relationship between meteorological factors and ET0, and the model shows good generalizability in Xinjiang, which can provide new methods for daily ET0 prediction in Xinjiang.

Key words: Xinjiang;particle swarm optimization;extreme learning machine;reference crop evapotranspiration;model accuracy

参考作物蒸散量(ET0)是水循环研究中的重要组成部分,也是优化农业用水的重要变量,在水资源可持续管理及农业精准灌溉中起着重要作用[1]。在农业生态系统中,约2/3的降水量由作物蒸散过程损失[2-3]。由于频繁的干旱及农业、个人和工业用户之间对水资源的竞争,目前的农业生产用水量已经减少[4]。……

登录APP查看全文

猜你喜欢

新疆
走进新疆
在新疆(四首)
新疆和静高级中学简介
新疆反恐,暗流与真相
怀念新疆
怀念新疆
今非昔比的新疆“巴扎”
「一帯一路」の砂漠の要衝——新疆カシュガルを行く
新疆多怪
新疆对外开放山峰