Abstract

Global warming and climate change are significant contributors to water scarcity, necessitating the efficient management of water resources for long-term sustainability. Agriculture, as one of the largest consumers of water, plays a critical role in this scenario. However, the water used in agricultural activities is often polluted and unsuitable for reuse. Thus, optimizing water management systems in agricultural irrigation is essential. This paper introduces an automated crop irrigation system that leverages real-time soil sensor readings to predict water treatment plans. Additionally, the system integrates weather conditions into its decision-making process before initiating water supply. A decision-making function has been developed to predict water treatment requirements and future weather conditions accurately. To achieve this, an Artificial Neural Network (ANN) algorithm has been trained to perform dual predictions.
To enhance the effectiveness of the dataset used for training the machine learning model, a preprocessing algorithm for soil moisture sensor data has been proposed. Experiments and simulations were conducted to evaluate the system’s performance. The results demonstrate that the proposed ANN-based weather prediction technique achieves a higher accuracy of 99.4% compared to Support Vector Machine (SVM), which achieves 95% accuracy. For soil condition predictions, the system delivers an accuracy of 88.4%. Moreover, the system’s decision-making and training times were assessed, showing that decisions can be made within fractions of a second. The training process, conducted on location-specific data, requires minimal time and only needs to be performed once for a given location. These findings highlight the efficiency and practicality of the proposed automated irrigation system in optimizing water management in agriculture.