| Cite this article: | 秦安振,曹睿喆.基于时间序列与机器学习的参考作物蒸散量预测研究[J].灌溉排水学报,0,():-. |
| qinanzhen,CAO Ruizhe.基于时间序列与机器学习的参考作物蒸散量预测研究[J].灌溉排水学报,0,():-. |
|
| |
|
|
| This article has been：Browse 64Times Download 0Times |
|
|
|
| DOI： |
|
| Prediction of Reference Crop Evapotranspiration Based on Time Series and Machine Learning Models |
|
qinanzhen1, CAO Ruizhe2
|
|
1.Institute of Farmland Irrigation, CAAS;2.Xinxiang Hydrology and Water Resources Reporting Subcenter
|
| Abstract: |
| 【Background】Reference crop evapotranspiration (ET0) is a key factor for crop water requirement estimation.【Objective】To explore suitable ET0 prediction models, historical meteorological data in 2021-2022 and numerical weather forecast data in 2023 were used.【Method】The experiment was conducted in Xinxiang City, Henan Province, China. Prophet model, autoregressive integrated moving average (ARIMA) model, extreme learning machine (ELM) model, and the corresponding hybrid prediction models were established for ET0 forecast. The predicted values were compared with FAO-56 Penman Monteith model.【Result】Meteorological parameters, including maximum temperature, minimum temperature, solar radiation, and sunshine hours, had significant correlation with ET0 and can be input factors for the models. Prophet and ARIMA models well predicted the periodic changes of annual ET0. However, when ET0 values were ≥ 6.5 mm/d, there existed significant errors. ELM model well fitted the complex nonlinear features of ET0, with an R2 value 11% higher than time series models. ELM-ARIMA hybrid model performed best in 1-10 d ET0 forecast. The MAE, RMSE and MBE of the ELM-ARIMA model were 64.5%, 72.9% and 65.6% lower than those of the single models. The R value was 12.9% higher. ELM-ARIMA hybrid model had the highest correlation between the predicted and observed ET0 values, with an R2 of 0.945. 【Conclusion】It was concluded that ELM-ARIMA hybrid model could be a reliable ET0 prediction model for the North Henan Province. |
| Key words: numerical weather prediction; hybrid model; Prophet model; autoregressive and moving average model |
|
|
|
|
|
|
