STM32-based IoT Soil Moisture Monitoring System with Deep Learning–Powered Predictive Irrigation Control using CNN-LSTM

A high-performance and smart irrigation system is necessary for sustainable farming, whereby traditional soil moisture monitoring systems are usually limited in accuracy, timely responsiveness, and lack of flexibility to changing environmental circumstances. To overcome these shortcomings, this work proposes an STM32-based IoT multifunctional Soil Moisture Monitoring System integrated with a hybrid Convolutional Neural Network with Long Short-Term Memory (CNN-LSTM) algorithm for predictive irrigation control. The microcontroller used, STM32, reads the real-time soil moisture of the field sensors, whereas the CNN reads spatial patterns, and the LSTM reads temporal relations, which are used to predict future moisture levels. The system is autonomous, as it regulates irrigation to reduce water wastage based on predictions. The proposed algorithm was experimentally evaluated and proved to have a higher prediction accuracy, shorter response time, and higher water-use efficiency than the traditional threshold-based and linear regression models. The general system provides low power usage, strong performance, improved accuracy, and is applicable in smart farming, automation of greenhouses, and large-scale agricultural Internet of Things applications.