Long-term performance validation of NH₃ concentration prediction model for virtual sensor application in livestock...

Area of Science:

• Agricultural Engineering
• Environmental Monitoring
• Sensor Technology

Background:

• Ammonia (NH3) is a hazardous gas emitted from livestock facilities, posing risks to animal health and requiring effective monitoring.
• Traditional NH3 sensors face durability issues due to the gas's corrosive nature, complicating continuous monitoring in agricultural environments.
• Accurate NH3 concentration data is vital for maintaining optimal livestock housing conditions and environmental compliance.

Purpose of the Study:

• To propose and validate a virtual sensor concept for ammonia (NH3) concentration monitoring in livestock facilities.
• To assess the long-term performance and reliability of a data-driven NH3 prediction model.
• To identify challenges and requirements for implementing virtual sensing in dynamic livestock environments.

Main Methods:

• Development of a data-driven model to predict ammonia (NH3) concentrations.
• Long-term performance validation of the prediction model under real-world livestock facility conditions.
• Analysis of model performance variations based on changes in data generation patterns and training data periods.

Main Results:

• The NH3 prediction model's performance significantly degrades when data patterns change due to external factors (weather) or internal operations.
• Model accuracy is sensitive to the duration of training data used for updates, indicating a need for adaptive strategies.
• The virtual sensor approach shows potential but requires robust versioning and update management for sustained effectiveness.

Conclusions:

• A virtual sensor system, supported by a data-driven model, can complement physical ammonia sensors in livestock facilities.
• Effective management, including versioning and adaptive updates, is essential for maintaining virtual sensor accuracy in response to environmental and operational shifts.
• The proposed virtual sensor concept offers a pathway to enhance NH3 monitoring efficiency and reduce operational costs in livestock management.