Forecasting method of water requirement of winter wheat

doi:10.6048/j.issn.1001-4330.2024.07.004

Abstract

Abstract:

【Objective】Based on the meteorological data related to water demand forecasting of winter wheat, a water demand forecasting model with fewer parameters was constructed to improve the robustness of water demand forecasting,provides a more reliable method for forecasting water demand based on meteorological information.【Methods】Meteorological data of Qitai County in recent five years were selected, and the water requirement of winter wheat calculated by Penman-Monteith formula was approximately the real water requirement. Four variables including average temperature, wind speed, humidity and precipitation were taken as input parameters. The water requirement of winter wheat was forecasted, and the prediction of CNN-BiLSTM was compared with that of LSTM, BiLSTM and other 6 models. 【Results】The results showed that when a few parameters were fed into BP, RNN, LSTM, improved BiLSTM and CNN-BiLSTM models to predict water demand, the prediction effect of BP neural network was poor. In the model evaluation, CNN-BiLSTM showed an R² improvement of about 14% over LSTM and a MSE reduction of about 3.8. 【Conclusion】CNN-BiLSTM model is more accurate in predicting wheat water demand.

Key words: winter wheat; water demand; forecast; LSTM; CNN-BiLSTM

摘要：

【目的】构建冬小麦需水量预测模型,提高需水量预测的精准度,为基于气象信息的需水量预测提供更为可靠的方法。【方法】选取新疆奇台县近5年的气象数据,采用公式Penman-Monteith计算冬小麦需水量(近似为真实需水量),基于CNN-BiLSTM模型,将平均温度、风速、湿度和降水量4个变量作为输入参数,预测冬小麦需水量,对比评估预测CNN-BiLSTM与LSTM、BiLSTM等6种模型的精准性。【结果】采用少量参数分别输入BP、RNN、LSTM、改进的BiLSTM和CNN-BiLSTM等模型中预测需水量,BP神经网络的预测效果较差。在模型评估中,CNN-BiLSTM比LSTM的R²提高约8%,MSE降低约0.56。【结论】CNN-BiLSTM模型对小麦需水量预测更加精准。

关键词: 冬小麦, 需水量, 预测, LSTM, CNN-BiLSTM

CLC Number:

S512

DU Yun, ZHANG Jingjing, LEI Jiacheng, LI Bo, LI Yongfu. Forecasting method of water requirement of winter wheat[J]. Xinjiang Agricultural Sciences, 2024, 61(7): 1590-1596.

杜云, 张婧婧, 雷嘉诚, 李博, 李永福. 冬小麦需水量的预测模型对比分析[J]. 新疆农业科学, 2024, 61(7): 1590-1596.

Figures/Tables 8

References 20

[1]	苏芳蕊. 小麦群体生长可视化系统的设计与实现[D]. 郑州: 河南农业大学, 2011.
	SU Fangrui. Design and Implementation of the Growing Visualization System for Wheat Population[D]. Zhengzhou: Henan Agricultural University, 2011.
[2]	2021年度《中国水资源公报》发布[J]. 水资源开发与管理, 2022, 8(7):85.
	China Water Resources Bulletin for 2021[J]. Water Resources Development and Management, 2022, 8(7):85.
[3]	Qin H H, Cai X M, Zheng C M. Water demand predictions for megacities: system dynamics modeling and implications[J]. Water Policy, 2018, 20(1): 53-76.
[4]	邓忠, 翟国亮, 宗洁, 等. 微灌系统堵塞机理分析与微灌过滤器研究进展[J]. 节水灌溉, 2014, (8): 71-74.
	DENG Zhong, ZHAI Guoliang, ZONG Jie, et al. Clogging mechanism analysis of micro-irrigation system and the advance of research on micro-irrigation filter[J]. Water Saving Irrigation, 2014, (8): 71-74.
[5]	胡程达, 方文松, 王红振, 等. 河南省冬小麦农田蒸散和作物系数[J]. 生态学杂志, 2020, 39(9): 3004-3010.
	HU Chengda, FANG Wensong, WANG Hongzhen, et al. Evapotranspiration and crop coefficient of winter wheat cropland in Henan Province[J]. Chinese Journal of Ecology, 2020, 39(9): 3004-3010.
[6]	董楠, 吕新, 侯振安, 等. 基于彭曼公式的膜下滴灌棉田灌水量研究[J]. 新疆农业科学, 2012, 49(4): 617-624.
	DONG Nan, LYU Xin, HOU Zhenan, et al. Study on the irrigation water amount under film mulch drip irrigation according to penman-monteith formula[J]. Xinjiang Agricultural Sciences, 2012, 49(4): 617-624.
[7]	白桦, 鲁向晖, 杨筱筱, 等. 基于彭曼公式日均值时序分析的中国蒸发能力动态成因[J]. 农业机械学报, 2019, 50(1): 235-244.
	BAI Hua, LU Xianghui, YANG Xiaoxiao, et al. Attribution analysis on changes in evaporation capacity based on mean diurnal time-series analysis of penman equation in China[J]. Transactions of the Chinese Society for Agricultural Machinery, 2019, 50(1): 235-244.
[8]	李志新, 赖志琴, 龙云墨. 基于GA-Elman神经网络的参考作物需水量预测[J]. 节水灌溉, 2019, (2):117-120.
	LI Zhixin, LAI Zhiqin, LONG Yunmo. Prediction of reference crop water requirement based on GA-Elman neural network[J]. Water Saving Irrigation, 2019, (2):117-120.
[9]	刘洪山, 王卫星, 孙道宗, 等. 基于GA-BP神经网络的果园需水量预测[J]. 排灌机械工程学报, 2020, 38(12): 1258-1263.
	LIU Hongshan, WANG Weixing, SUN Daozong, et al. Prediction model of orchard water requirement based on BP neural network and genetic algorithm[J]. Journal of Drainage and Irrigation Machinery Engineering, 2020, 38(12): 1258-1263.
[10]	刘婧然, 刘心, 武海霞, 等. 基于GA优化的支持向量机模型在青椒作物需水量预测中的应用[J]. 节水灌溉, 2021, (1): 70-76.
	LIU Jingran, LIU Xin, WU Haixia, et al. Application of support vector machine model based on GA optimization in water consumption prediction of green peppers[J]. Water Saving Irrigation, 2021, (1): 70-76.
[11]	邓皓, 李文竹, 刘婧然, 等. 基于MIV-MEA-Elman神经网络的核桃果实膨大期需水量预测[J]. 节水灌溉, 2020, (4): 68-72.
	DENG Hao, LI Wenzhu, LIU Jingran, et al. Water demand prediction for walnut fruit during swelling period based on MIV-MEA-elman neural network[J]. Water Saving Irrigation, 2020, (4): 68-72.
[12]	夏泽豪, 翁绍捷, 罗微, 等. 基于灰色神经网络的作物需水量预测模型研究[J]. 中国农机化学报, 2015, 36(2): 219-223.
	XIA Zehao, WENG Shaojie, LUO Wei, et al. Crop water requirement forecasting model based on grey neural network[J]. Journal of Chinese Agricultural Mechanization, 2015, 36(2): 219-223.
[13]	王景雷, 康绍忠, 孙景生, 等. 基于贝叶斯最大熵和多源数据的作物需水量空间预测[J]. 农业工程学报, 2017, 33(9): 99-106, 315.
	WANG Jinglei, KANG Shaozhong, SUN Jingsheng, et al. Spatial prediction of crop water requirement based on Bayesian maximum entropy and multi-source data[J]. Transactions of the Chinese Society of Agricultural Engineering, 2017, 33(9): 99-106, 315.
[14]	赵实, 赵雷, 周玉良. 基于彭曼公式的宿州水面蒸发计算与分析[J]. 安徽农业科学, 2022, 50(23): 199-202.
	ZHAO Shi, ZHAO Lei, ZHOU Yuliang. Analysis of evaporation based on penman equation in Suzhou city[J]. Journal of Anhui Agricultural Sciences, 2022, 50(23): 199-202.
[15]	田旭浪. 伊犁河灌区滴灌小麦作物系数及灌溉制度优化研究[D]. 石河子: 石河子大学, 2022.
	TIAN Xulang. Optimization of wheat crop coefficient and irrigation system under drip irrigation in Yili River irrigation district[D]. Shihezi: Shihezi University, 2022.
[16]	崔丽珍, 张清宇, 郭倩倩, 等. 基于CNN-LSTM的井下人员行为模式识别模型[J/OL]. 无线电工程:1-9 [2023-04-24].
	CUI Lizhen, ZHANG Qingyu, GUO Qianqian, et al. Behavior pattern recognition model of downhole personnel based on CNN-LSTM[J/OL]. Radio Engineering: 1-9 [2023-04-24].
[17]	梁燕华, 沈奋博, 谢子殿, 等. 基于LSTM模型的冲击地压预测方法研究[J/OL]. 中国矿业:1-9 [2023-04-16].
	LIANG Yanhua, SHEN Fenbo, XIE Zidian, et al. Research on the prediction method of rock burst based on LSTM model[J/OL]. China Mining Industry:1-9 [2023-04-16].
[18]	余琼芳, 牛冬阳. 基于LSTM网络的矿山压力时空混合预测[J]. 电子科技, 2023, 36(2): 67-72.
	YU Qiongfang, NIU Dongyang. Mixed prediction of mine pressure time and space based on LSTM network[J]. Electronic Science and Technology, 2023, 36(2): 67-72.
[19]	刘泽, 张闯, 齐磊, 等. 基于CNN-BiLSTM的锂电池剩余使用寿命概率密度预测[J]. 电源技术, 2023, 47(1): 57-61.
	LIU Ze, ZHANG Chuang, QI Lei, et al. Prediction of probability density of remaining useful life of lithium ion battery based on CNN-BiLSTM[J]. Chinese Journal of Power Sources, 2023, 47(1): 57-61.
[20]	王欢, 李鹏, 曹敏, 等. 基于CNN_BiLSTM的长短期电力负荷预测方法[J]. 计算机仿真, 2022, 39(3): 96-103.
	WANG Huan, LI Peng, CAO Min, et al. Long-term and short-term power load online prediction method based on CNN_BiLSTM[J]. Computer Simulation, 2022, 39(3): 96-103.

日均温度 Average daily temperature (℃)	风速 Wind speed	相对湿度 Relative humidity	降水量 Precipitati on	小麦需水量 Wheat water requirement (mm)
22. 05	7	46	0	4.62
22. 05	2	62	0	3.41
21.2	2	21	0	8.08
20. 85	1	75	0	1.7
24.5	2	54	5	4.29
…	…	…	…	…
5.05	2	72	0	0.92
4.8	2	46	0	2.05
2.75	2	28	0	3.07
1.8	2	30	0	2.74
1.05	0	51	0	0.45

日均温度 Average daily temperature (℃)	风速 Wind speed	相对湿度 Relative humidity	降水量 Precipitati on	小麦需水量 Wheat water requirement (mm)
22. 05	7	46	0	4.62
22. 05	2	62	0	3.41
21.2	2	21	0	8.08
20. 85	1	75	0	1.7
24.5	2	54	5	4.29
…	…	…	…	…
5.05	2	72	0	0.92
4.8	2	46	0	2.05
2.75	2	28	0	3.07
1.8	2	30	0	2.74
1.05	0	51	0	0.45

参数 Parameter	参数值 Parameter value
卷积层 Convolution layer	卷积核个数: 128 卷积核维度:1维
池化层 Pooled layer	最大池化窗口大小: 1
BILSTM层 BiLSTM layer	神经元个数64
训练参数设置 Training par ameter setting	Dropout随即失活: 0.1 优化器: Adan Epoch: 100

参数 Parameter	参数值 Parameter value
卷积层 Convolution layer	卷积核个数: 128 卷积核维度:1维
池化层 Pooled layer	最大池化窗口大小: 1
BILSTM层 BiLSTM layer	神经元个数64
训练参数设置 Training par ameter setting	Dropout随即失活: 0.1 优化器: Adan Epoch: 100

模型 Model	R²	MAE	MSE	RMSE
BP神经网络	0.803	0.932	1.347	1.16
RNN	0.7958	0.933	1.394	1.181
单层LSTM	0.8496	0.793	1.047	1.023
两层LSTM	0.889 4	0.606	0.782	0.884
三层LSTM	0.867 7	0.575	0.853	0.924
BiLSTM	0.890 1	0.593	0.727	0.853
CNN-BiLSTM	0.928 9	0.4	0.486	0.687