| 引用本文: | 李润童,邢立文,崔宁博,等.基于群智能算法优化LSTM模型的参考作物蒸散量模拟研究[J].灌溉排水学报,2026,45(2):20-30. |
| LI Runtong,XING Liwen,CUI Ningbo,et al.基于群智能算法优化LSTM模型的参考作物蒸散量模拟研究[J].灌溉排水学报,2026,45(2):20-30. |
|
| 摘要: |
| 【目的】基于有限气象资料实现西北干旱地区逐日参考作物蒸散量（ET0）高精度模拟。【方法】将西北干旱地区划分为4个亚气候区（温带大陆性干旱区、温带大陆性高温干旱区、高原大陆性半干旱区和温带季风半干旱区），选取8个代表性气象站点1961—2019年逐日气象数据作为输入参数，以FAO-56 Penman-Monteith模型计算的ET0作为标准值。采用遗传算法（GA）和粒子群优化算法（PSO）对长短期记忆网络（LSTM）超参数进行优化，构建了LSTM、GA-LSTM和PSO-LSTM共3种深度学习模型。针对西北干旱地区气象数据缺乏的情况，设计了3种输入组合方案（温度-辐射型、温度型、温度-湿度型），仅利用气温、日照时间和相对湿度等基础气象要素，建立了9种模型组合。与Priestley-Taylor、Hargreaves-Samani和Romanenko 3种经验模型对比，评估深度学习模型的ET0模拟精度。【结果】PSO-LSTM模型模拟精度最高，决定系数（R2）、Nash-Sutcliffe系数（NSE）、均方根误差（RMSE）、相对均方根误差（RRMSE）、平均绝对误差（MAE）和综合性指标（GPI）分别为0.831~0.923、0.801~0.922、0.476~0.866 mm/d、0.190~0.382、0.299~0.627 mm/d和0.208~0.598，其中，温度-辐射型PSO-LSTM1模型在4个气候分区的ET0模拟精度最高，R2达0.893~0.923；智能优化算法可显著提升LSTM模型性能，且PSO算法的提升效果优于GA算法。【结论】基于温度-辐射型输入策略的PSO-LSTM模型在西北干旱地区ET0模拟中表现最优，为西北干旱地区及类似气候区域ET0准确模拟提供了有效方法。 |
| 关键词: 神经网络；遗传算法；粒子群优化算法；ET0模拟；西北干旱地区 |
| DOI：10.13522/j.cnki.ggps.2025284 |
| 分类号: |
| 基金项目: |
|
| A proposed model for simulating daily reference crop evapotranspiration |
|
LI Runtong, XING Liwen, CUI Ningbo, JIANG Shouzheng, WANG Zhihui,
ZHU Guoyu, LIU Jincheng, HE Qingyan
|
|
1. State Key Laboratory of Hydraulics and Mountain River Engineering & College of Water Resources and HydroPower,
Sichuan University, Chengdu 610065, China; 2. Daying County Bureau of Agriculture and Rural Affairs, Suining 629399, China;
3. Sichuan Academy of Agricultural Machinery Sciences, Chengdu 610066, China
|
| Abstract: |
| 【Objective】The reference crop evapotranspiration (ET0) is an important parameter for irrigation scheduling and water resource management, but its estimation is often constrained by limited meteorological data. This paper proposes a method to bridge this technological gap.【Method】The study was conducted for the arid regions in Northwestern China. We divided the regions into four climatic zones: temperate continental arid zone, temperate continental high-temperature arid zone, plateau continental semi-arid zone, and temperate monsoon semi-arid zone. Daily meteorological data from 1961 to 2019 at eight representative weather stations in the region were used as baseline data for the model, and the ET0 calculated using the FAO-56 Penman-Monteith model served as the benchmark ET0. Three deep learning models, LSTM, GA-LSTM and PSO-LSTM, optimized the long short-term memory (LSTM) network hyperparameters using a genetic algorithm (GA) and particle swarm optimization (PSO), were developed to estimate ET0 by using air temperature, sunshine duration and relative humidity, either separately or in their combination, as the independent variables. We also compared the proposed models with three other empirical models: Priestley-Taylor, Hargreaves-Samani, and Romanenko model.【Result】The PSO-LSTM model was the most accurate, with its coefficient of determination (R2) and Nash-Sutcliffe efficiency (NSE) being 0.831-0.923 and 0.801-0.922, and RMSE, RRMSE, MAE and GPI being 0.476-0.866 mm/d, 0.190-0.382, 0.299-0.627 mm/d and 0.208-0.598, respectively. For areas with only temperature and radiation data, the PSO-LSTM1 model was the most accurate for all four climatic zones, with its R2 varying from 0.893 to 0.923. Intelligent optimization significantly improved the accuracy of LSTM, especially with PSO.【Conclusion】The PSO-LSTM model using temperature and radiation was found to be the most effective and reliable for estimating ET0 in arid regions in Northwestern China. It can be used to estimate ET0 in regions where meteorological data are limited. |
| Key words: neural networks; genetic algorithm; particle swarm optimization; ET0 simulation; arid regions in Northwestern China |
