物理プロセスを組み込んだ深層学習による水田内温度環境の予測モデルの構築
| Project/Area Number |
22KJ0781
|
|---|
| Project/Area Number (Other) |
22J10619 (2022)
|
|---|
| Research Category |
Grant-in-Aid for JSPS Fellows
|
| Allocation Type | Multi-year Fund (2023)
Single-year Grants (2022) |
|---|
| Section | 国内 |
|---|
| Review Section |
Basic Section 41030:Rural environmental engineering and planning-related
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
謝 文鵬 東京大学, 生産技術研究所, 特別研究員(PD)
|
|---|
| Project Period (FY) |
2023-03-08 – 2024-03-31
|
| Project Status |
Completed (Fiscal Year 2023)
|
| Budget Amount *help |
¥1,700,000 (Direct Cost: ¥1,700,000)
Fiscal Year 2023: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 2022: ¥900,000 (Direct Cost: ¥900,000)
|
|---|
| Keywords | paddy field / PGNN / water temperature / 水田水温 / 熱収支モデル |
|---|
| Outline of Research at the Start |
本研究では、 観測対象水田において取得した観測データに基づき、物理プロセスを組み込んだニューラルネットワークを構築し、水田内の温度環境を再現する機械学習モデルを開発する。また、開発した機械学習モデルに対して、計算精度や他地域への適用性の評価、感度分析を実施する。さらに機械学習モデルの決定プロセスを局所的な解釈可能性とグローバルな解釈可能性の観点から明らかにすることで、モデルの信頼性を評価する。最後に、気象予測と組み合わせて水田内の温度環境を予測することで、水稲の品質向上に向けた水管理を支援するツールのプロトタイプを提示する。
|
| Outline of Annual Research Achievements |
Firstly, a paper titled "Interpretable Framework of Physics-guided Neural Network with Attention Mechanism: Simulating Paddy Field Water Temperature Variations" was published in the esteemed journal WRR. This work elucidates the interpretability of a novel neural network framework integrated with physics guidance and attention mechanisms, specifically applied to simulate variations in paddy field water temperature.
Furthermore, another paper was published in 2022 in the journal Irrigation and Drainage, titled "Development of a short-term hybrid forecast model of paddy water temperature as an alert system for high-temperature damage." This research focuses on the development of a hybrid forecasting model tailored for predicting short-term paddy water temperature variations, crucial for mitigating high-temperature damage in agricultural settings.
Additionally, active participation was observed in the AGU conferences of 2022 and 2023, where presentations were delivered to showcase the latest advancements in the research domain.
In summary, over the course of these two years, significant insights have been gained into the integration of deep learning and physical models, along with an understanding of the advantages that deep learning offers for modeling natural phenomena. This research has notably enhanced the understanding and prediction capabilities of dynamic variations in paddy field temperatures. Moreover, valuable insights have been derived for informing agricultural irrigation systems, policy support, and guiding recommendations aimed at optimizing agricultural yields.
|
Report
(2 results)
Research Products
(7 results)
