基于DBO-LSTM-Attention模型的设施番茄茎粗预测OA

Stem diameter variation is a vital physiological indicator for evaluating tomato growth status.Accurately predicting its dynamic changes in advance is of great significance for precise regulation of greenhouse environments.Aiming at the deficiencies of existing prediction models such as insufficient feature extraction and weak capability in capturing long-term dependencies,this study proposed a stem diameter prediction model integrating dung beetle optimizer(DBO),long short-term memory(LSTM)network and attention mechanism,namely DBO-LSTM-Attention.The model used LSTM mod-ule to capture temporal dependencies between stem diameter dynamics and environmental factors including air temperature,air humidity,carbon dioxide concentration and photosynthetically active radiation.The attention mechanism was introduced to dynamically assign weights,thereby enhancing the model's focus on critical time steps.The DBO algorithm was adopted for adaptive hyperparame-ter optimization to improve model generalization perfor-mance.The results revealed that the DBO-LSTM-Attention model achieved superior prediction stability and accuracy in both short-term and long-term prediction tasks,with all evaluation indices outperforming those of the comparative models.Its performance declined slightly with the increase of prediction horizon,proving strong temporal modeling and generalization ability.In conclusion,the DBO-LSTM-Attention model can effectively fuse tomato plant growth parameters and greenhouse environmental factors to realize high-precision prediction of dynamic changes in stem diameter,and provide a theoretical reference for intelligent regulation of growing environments for protected tomatoes.