Continuous Multi-Zone Prediction Model to Monitor THI in Dairy Cattle Farms

Indoor environmental parameters have a significant influence on agricultural buildings, playing a crucial role in ensuring animal welfare and maintaining productivity. Effective prediction, along with continuous monitoring and controlling of near real-time parameters, can help manage energy consumption while optimizing thermal comfort levels for the herds. In this study, a multi-zone Temperature-Humidity Index (THI) prediction model was developed to support decision-making regarding environmental control across different zones within a dairy barn. The model was designed to serve as the foundation for future development of more advanced control systems. The NeuralProphet library was employed to build the predictive model, utilizing external variables such as temperature, relative humidity, and solar radiation. Validation of the predicted THI values against offline-calculated measurements confirmed the high accuracy of the model, with a mean error not exceeding four degrees. The integration of this THI prediction model into an innovative Smart Monitoring System (SMS) extended the functionality of an established workflow, which includes internal environmental data acquisition, data storage, and visualization. By dividing the barn into distinct zones and developing predictive models for each zone, the system provides more granular control over ventilation and thermal conditions. The developed model achieved high predictive accuracy for the three barn zones, with RMSE values of 3.86, 3.98, and 4.68, demonstrating its ability to capture spatial variations in indoor conditions. By incorporating this proactive predictive model within the SMS, farmers are empowered with real-time decision-making capabilities, seamlessly integrating data collection with actionable insights.