Multi-step prediction of greenhouse crop growth based on the SVR_Seq2Seq model
In greenhouse cultivation, ensuring optimal plant development relies upon precise forecasting of crop growth, which is a fundamental aspect of effective productivity. However, the challenge of obtaining accurate predictions over extended periods persists due to inherent complexities such as the infl...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-08-01
|
| Series: | Smart Agricultural Technology |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2772375525002199 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In greenhouse cultivation, ensuring optimal plant development relies upon precise forecasting of crop growth, which is a fundamental aspect of effective productivity. However, the challenge of obtaining accurate predictions over extended periods persists due to inherent complexities such as the influence of diverse environmental factors. The Support Vector Machine Sequence to Sequence (SVR_Seq2Seq) model introduced in this paper offers a multi-step prediction method for greenhouse crop growth. This approach integrates the Seq2Seq architecture with the Support Vector Regression model, enhancing the capacity of the latter to transfer memory information between multi-dimensional input sequences and predictive output sequences. Consequently, this integration improves the accuracy and stability of multi-step predictions. This study compares the SVR_Seq2Seq model with the LSTM model, Bi-LSTM model, Transformer model, and Seq2Seq model across different prediction steps. The results demonstrate that the greenhouse crop growth predictions from the SVR_Seq2Seq model are more consistent with the trend of the measured values. The multi-step crop growth prediction errors are <5 %, and the average MAE and the RMSE being 0.1540 and 0.1623 cm, respectively. Furthermore, we analyzed the universality of the SVR_Seq2Seq model for crop growth prediction of different plant varieties. The results show that the SVR_Seq2Seq model exhibits universal adaptability, and the minimum value of MAE for different crop datasets is 0.1500 cm, the maximum value is 0.1864 cm. It can be seen that the SVR_Seq2Seq model proposed offers a high prediction accuracy, stability, and generalizability, and can provide reliable data support for greenhouse crop growth health monitoring and yield management. |
|---|---|
| ISSN: | 2772-3755 |