Assessing the stability of indoor farming systems using data outlier detection

Area of Science:

• Agricultural Engineering
• Environmental Monitoring
• Sensor Technology

Background:

• Controlled Environment Agriculture (CEA) systems rely on accurate environmental data for stability and efficiency.
• Low-cost IoT sensors are increasingly used in CEA, necessitating data quality assessment.
• Lettuce cultivation trials were conducted at four different temperatures to evaluate sensor performance.

Purpose of the Study:

• To investigate the quality of air temperature data from low-cost IoT sensors in a small-scale CEA system.
• To assess CEA system stability by analyzing the correlation between cumulative agricultural operations (Agr.Ops) and air temperature data variability.
• To identify and analyze outliers in sensor data to understand their impact on system performance.

Main Methods:

• Collected air temperature data from IoT sensors throughout lettuce cultivation.
• Employed generalized linear model regression to analyze the relationship between cumulative Agr.Ops and z-scores of air temperature residuals.
• Utilized residual distribution and curve fitting to identify the best data distribution model (log-normal).

Main Results:

• A strong inverse relationship was found between cumulative Agr.Ops and residual z-scores, indicating decreased system stability with increased operations.
• Outliers in sensor data were linked to potential issues like sensor noise or drift.
• The system exhibited varying resilience to cumulative Agr.Ops across different trials.

Conclusions:

• The study highlights the importance of addressing data uncertainties in indoor farming through improved models, sensor selection, and redundancy.
• An alternative decomposition method effectively identified outliers and provided insights into system functionality.
• The findings offer a promising approach for enhanced monitoring and management of uncertainties in CEA systems.