Efficient Soil Temperature Profile Estimation for Thermoelectric Powered Sensors

Area of Science:

• Environmental Science
• Agricultural Technology
• Sensor Networks

Background:

• Powering Internet of Things (IoT) sensors for environmental and agricultural applications is a significant challenge.
• Exploiting temperature differences between air and soil presents a promising solution for energy harvesting.
• Accurate soil temperature profile data is crucial for effective energy-harvesting technologies.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting soil temperature profiles.
• To optimize energy harvesting for IoT sensors by improving soil temperature prediction accuracy and efficiency.
• To simplify the input parameters for soil temperature prediction models.

Main Methods:

• Utilized meteorological and soil temperature profile data from the Czech Republic.
• Trained machine learning models including Polynomial Regression (PR), Support Vector Regression (SVR), and Long Short-Term Memory (LSTM).
• Simplified model inputs to ambient temperature and solar irradiance.

Main Results:

• Achieved a prediction error of 0.79 °C for soil temperature profiles.
• Demonstrated a 10.9% reduction in temperature error compared to state-of-the-art studies.
• Significantly reduced computational costs by simplifying input parameters.

Conclusions:

• The proposed machine learning approach provides a more efficient and accurate method for predicting soil temperature.
• This advancement facilitates optimized energy harvesting for IoT sensors in environmental and agricultural settings.
• Simplified input parameters enhance the practicality and cost-effectiveness of the solution.