Research Finds Hybrid Model for Adaptive Temperature and Humidity Forecasting in Solar Greenhouses

Strategic Insights -

Researchers from Shenyang Agricultural University have developed a new RIME-optimized CNN-BiLSTM hybrid model for precise climate forecasting to protect agriculture. This model addresses the challenges of large diurnal temperature difference and extreme weather in Northeast China, which can significantly impact crop production. By using Pearson correlation analysis and the RIME algorithm, the model can automatically optimize hyperparameters to enhance spatiotemporal feature extraction and sequence learning. Experimental results show significant improvements in temperature and humidity prediction accuracy, with a 46% and 41% decrease in mean absolute error (MAE) and root mean square error (RMSE) respectively.

Key Takeaways:

  • The RIME-optimized CNN-BiLSTM hybrid model was developed to address the challenges of large diurnal temperature difference and extreme weather in Northeast China.
  • The model uses Pearson correlation analysis and the RIME algorithm to automatically optimize hyperparameters and enhance spatiotemporal feature extraction and sequence learning.
  • Experimental results show that the model has decreased the MAE and RMSE of temperature prediction by 46% and 41% respectively, and humidity prediction by 29% respectively.
  • The model has increased the R2 (coefficient of determination) of temperature and humidity by 4.4% and 1.8% respectively.
  • The research provides a data-driven and scalable framework for precise climate forecasting to protect agriculture.
  • The model was tested using a one-year dataset with 35,040 samples.
  • The researchers identified the key environmental variables that are crucial for intelligent climate regulation and sustainable crop production.

Statistics:

  • Temperature prediction accuracy: 46% decrease in MAE and 41% decrease in RMSE.
  • Humidity prediction accuracy: 29% decrease in MAE and 18% decrease in RMSE.
  • R2 (coefficient of determination) of temperature and humidity: 4.4% and 1.8% increase respectively.
  • Number of samples used in the one-year dataset: 35,040.
  • Number of environmental variables identified: Not specified.

Sources:

  • VerticalNews (2025, November 4). Research news report on Rime-cnn-bilstm for Data-driven Precision Agriculture: a Hybrid Model for Adaptive Temperature and Humidity Forecasting In Solar Greenhouses.
  • Rime-cnn-bilstm for Data-driven Precision Agriculture: a Hybrid Model for Adaptive Temperature and Humidity Forecasting In Solar Greenhouses. Computers and Electronics In Agriculture, 2025;238.
  • Shenyang Agricultural University, College of Engineering. "Rime-cnn-bilstm for Data-driven Precision Agriculture: a Hybrid Model for Adaptive Temperature and Humidity Forecasting In Solar Greenhouses." (2025).
  • Natural Science Foundation of Liaoning Province, Department of Education of Liaoning Province of China, Science and Technology Fund of Shenyang.