A Method to Predict CO₂ Mass Concentration in Sheep Barns Based on the RF-PSO-LSTM Model.

Area of Science:

• Agricultural Engineering
• Environmental Science
• Animal Science

Background:

• Elevated CO2 concentrations in sheep sheds negatively impact meat sheep health and growth.
• Accurate CO2 monitoring and early regulation are crucial for maintaining environmental safety and animal welfare in intensive farming operations.

Purpose of the Study:

• To develop and validate a novel prediction model for CO2 concentrations in sheep barns.
• To improve the accuracy and efficiency of CO2 monitoring for large-scale meat sheep farming.

Main Methods:

• Data preprocessing techniques including mean smoothing, linear interpolation, and normalization were applied to raw air quality data.
• A Random Forests (RF) algorithm was employed for feature selection, identifying light intensity, relative humidity, temperature, and PM2.5 as key predictors of CO2 concentration.
• A Long Short-Term Memory (LSTM) model, optimized via Particle Swarm Optimization (PSO) for hyperparameter tuning, was developed for CO2 prediction.

Main Results:

• The proposed RF-PSO-LSTM model achieved a Root Mean Square Error (RMSE) of 75.422 μg·m⁻³, a Mean Absolute Error (MAE) of 51.839 μg·m⁻³, and a Coefficient of Determination (R²) of 0.992.
• The model demonstrated a strong predictive performance, with its prediction curve closely aligning with the actual CO2 concentration data.
• Feature selection effectively reduced data dimensionality by identifying the most influential environmental parameters.

Conclusions:

• The RF-PSO-LSTM model provides a highly accurate and effective method for predicting CO2 concentrations in sheep farming environments.
• This predictive capability supports timely interventions for CO2 regulation, enhancing sheep welfare and farm environmental management.
• The integration of RF for feature selection and PSO for LSTM optimization offers a robust approach to environmental monitoring in livestock farming.