https://doi.org/10.1051/epjconf/202532801074
IoT Integrated Graphene Derivative-Based Capacitive Sensors with Machine Learning Classification for Precision Soil Moisture Monitoring Process
1 Ph. D. Scholar, Lovely Professional University, Punjab, India
2 SEEE, Lovely Professional University, Punjab, India
3 Ph. D. Scholar, Lovely Professional University, Punjab, India
* Corresponding author: neema.ukani@gmail.com
Published online: 18 June 2025
Soaring demand for sustainable agriculture critically requires the timely and precise monitoring of soil moisture to optimize irrigation applications and save scarce water resources. While conventional soil moisture sensing methods such as gravimetric analysis and neutron probes are accurate, they possess severe limitations including high cost, labor intensity, and unsuitability for large-scale field deployment. Also, the existing electronic sensors do not possess the desired sensitivity, reliability, and scalability needed for modern precision agriculture processes. To address these problems, a novel low-cost, IoT-enabled soil moisture sensing system based on GO and rGO capacitive sensors is presented. Nanomaterials were selected on account of superior electrical conductivity, large surface area, and tunable functional groups that allow for larger sensitivity changes in moisture variations. The sensors were fabricated using drop-casting techniques at three different concentrations (0.1 mg/ml, 1 mg/ml, 10 mg/ml) on interdigitated electrode arrays which were experimentally validated over a range of gravimetric water contents (1-17%). The sensing system featured a custom capacitance-to-frequency conversion circuit and ESP8266-based wireless data transmission module for real-time cloud integration via MATLAB ThingSpeak. Machine learning techniques were integrated for classifying soil moisture levels. Principal Component Analysis (PCA) was employed for dimensionality reduction and was able to capture more than 92% and 95% of data variance for the GO and rGO sensors, respectively. Afterward, the k-means clustering method, supported by the elbow method, enabled our correct classification into dry, moderate, and wet moisture levels with silhouette scores of 0.88 (GO) and 0.91 (rGO) Sets. This work demonstrates a strong, scalable, and data-centric sensing solution which potentially provides intelligent irrigation management in the precision agriculture process in real-time scenarios.
© The Authors, published by EDP Sciences, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
