Grid Search for Lowest Root Mean Squared Error in Predicting Optimal Sensor Location in Protected Cultivation Systems

Area of Science:

• Agricultural Engineering
• Environmental Monitoring
• Machine Learning Applications

Background:

• Inconsistent internal climates in protected cultivation systems hinder optimal yield due to inadequate environmental monitoring and control.
• Strategic sensor placement is crucial for effective indoor environment management and potential operational cost reduction.

Purpose of the Study:

• To determine the optimal number and placement of sensors in protected cultivation systems using a multi-objective, supervised machine learning approach.
• To evaluate the impact of environmental fluctuations on sensor requirements over different time scales (daily, weekly, monthly).

Main Methods:

• A gradient boosting algorithm, a tree-based model, was employed to analyze time-series data of measured (temperature, humidity) and derived (dew point, enthalpy) environmental conditions.
• Training and validation data were maintained in a time-dependent sequence to preserve feature relationships.
• Sensor variations were assessed across different temporal resolutions to understand environmental impact on optimal sensor configuration.

Main Results:

• Fewer than 32% of the 56 evaluated sensors were sufficient for optimal monitoring of the protected cultivation environment.
• Derived environmental properties provided a more accurate representation of indoor air conditions compared to directly measured parameters, enhancing model performance.
• Sensor requirements varied, with the highest need observed in May, indicating seasonal environmental influence.

Conclusions:

• A machine learning model can effectively determine the optimal number and positions for sensors in protected cultivation.
• Optimized sensor networks reduce hardware costs while ensuring precise environmental control for improved crop production.
• Utilizing derived atmospheric properties alongside direct measurements improves the accuracy of environmental monitoring models.