Identifying Uncertainties In Air Temperature Data Of An Indoor Farming System
Description
A small growth chamber was used to grow lettuce for 5 weeks during four air temperature trials, and an automated control system was used to control the environmental conditions of the air and root zone for the plants. A sensor array made up of low-cost arduino connected sensors would collect aerial (air temperature, relative humidity, CO₂ concentration) and root-zone data (water temperature, pH, EC, DO) and control the hydroponic system and carbon dioxide enrichment, while a reference sensor was used to collect aerial environmental conditions (air temperature, relative humidity, CO₂ concentration) to compare with the low-cost data sets. We hypothesized that our alternative decomposition method would successfully identify uncertainty occurrences in the data collected throughout this experiment since this data had many gaps in data when the data collection system would stop functioning or for other uncertainties. The small size of the growth chamber would also make any agricultural operations (any actions where humans would enter/exit and be inside the small chamber for any time), sensor failures, or other such uncertainties have a significant impact on system operations and reliability, making this decomposition method necessary for data quality control. Only the air temperature data from the low-cost and reference sensors was used to test the alternative decomposition method since the standard decomposition methods failed to successfully de-seasonalize the data.
Files
Steps to reproduce
Environmental growth chamber (Environmental Growth Chambers, TC2, Chagrin Falls, Ohio) of 2.74 m × 2.74 m × 2.39 m (width × depth × height) controlled the temperature, relative humidity (RH), and airflow. CO2 enrichment : constant 800 mg/L. A day/night interval of 16.5/7.5 h was used for both lighting and temperature control. Relative humidity was set to a maximum of 60% RH with no minimum. Five MEGA2560 boards (Elegoo, MEGA 2560, Shenzen, China) were used for the environmental monitoring and control, and one ESP32 (Espressif Systems, ESP32-S2-Saola-1, Shanghai, China) board was used for agricultural operations monitoring and connected to a single computer tower that acted as a data logger by logging data online to a MySQL database, and locally as a .txt file using Python scripts. All data was timestamped in Python. LC sensor data was collected at 10 sec intervals, while the reference system collected data at 5 second intervals. Unique parameters and duplicate data sets were averaged per minute. Air temperature, RH, and CO2 data were collected at the two plant canopy levels to ensure the operation of the system matched the needs of the plants. The reference sensor was placed away from the growing system to monitor the overall chamber conditions. LC sensors were placed next to the reference sensor to compare data gathered in identical conditions. A photocell monitored door opening/closing. The sensors used were as followed : reference sensor (Hydrofarm, APCEM2, Petaluma, CA), float switches (Anndason, DP3500, Shenzen, China), dosing pumps (Atlas Scientific, EZOTM-PMP, Long Island City, NY), CO₂ concentration (Vaisala, GMP252, Vantaa, Finland), air temperature and relative humidity sensors (Adafruit, SHT30, New York, NY), multiplexer for SHT30 (Adafruit, TCA9548A, New York, NY) nutrient solution temperatures (Aideepen, DS18B20, Shenzen, China), photoresistor (Adafruit, CdS photocell, New York, NY), pH (Atlas Scientific, #ENV-40-pH, Long Island City, NY), electrical conductivity (Atlas Scientific, #ENV-40-EC-K1.0, Long Island City, NY), and dissolved oxygen (Atlas Scientific, #ENV-40-DOX, Long Island City, NY), tentacle shield (Whitebox, T1.16, Schlatt TG, Switzerland), electrical consumption (CrocSee, CRS-022B, Shenzen, China).