Predictive Water Supply Optimization Using Sensor Data, Machine Learning, and Weather Forecasts

Global warming and climate change are key contributors to water scarcity. Consequently, judicious management of water resources is essential for enduring sustainability. Agriculture is a significant consumer of water among several large-scale industries. The water utilised in agriculture has become polluted and is unfit for usage. Therefore, enhancing water management techniques for agricultural irrigation is imperative. This work introduces an automated irrigation system that employs real-time soil sensor data to predict the water management approach. Conversely, the decision-making function has been refined, and the weather conditions have been evaluated prior to initiating the water supply. The decision-making function requires anticipating the water treatment strategy and forthcoming weather conditions, and the Artificial Neural Networks (ANN) system has been trained to effectively predict both variables. A pre-processing strategy for the soil moisture sensor dataset has been proposed to improve its efficacy for training with machine learning algorithms. The assessments have been executed and a simulation has been carried out. The performance results demonstrate that the proposed weather prediction method employing ANN attains more accuracy compared to Support Vector Machines (SVM). Additionally, the period for training and decision-making of the system has been evaluated, indicating that choices may be made in a fraction of a second. The training period is considered sufficient, necessitating only one session of training on location-specific data.