Water optimization technique for precision irrigation system using IoT and machine learning

• Main Authors: Maria Manuel Vianny, D., John, A., Kumar Mohan, S., Sarlan, A., Adimoolam,, Ahmadian, A.
• Format: Article
• Published: Elsevier Ltd 2022
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85131445161&doi=10.1016%2fj.seta.2022.102307&partnerID=40&md5=e76b340a794685a1d01323c5b0261c4f; http://eprints.utp.edu.my/33349/

Water management system and energy optimization are most important for today's agriculture. Precision irrigation manages the water utilization and reduces the utilization of energy. The irrigation system consists of observation system, storing, processing and it helps in decision making as well. Internet of Things (IoT) components is used to observe the various information's, such as soil moisture, soil temperature, weather conditions, and environmental conditions from the fields. With the help of various information, the proposed work optimizes the requirements and thus reduces energy. The cloud environment is used to store the observed parameters from the IoT components. The machine learning algorithm is useful in taking the process of data, from the cloud environments, and do forecasting on the irrigation system. In this work, proposed a hybrid model for irrigation system and do the decision making using, different observation of parameters. This work consists of K-Nearest Neighbors (KNN), Gradient Boosting-based Tree (GBT), Long Short-Term Memory (LSTM) and Spearman's rank correlation. The KNN is used to gather the nearest sensing information, with the help of various parameters. The GBT is used to predict the real values based on the observations. The LSTM is used to predict, time series prediction using different observations of the surrounding. The Spearman's rank correlation is used to receive the correlated values of real, and time series prediction. Using the GBT, LSTM and spearman rank correlation, the consolidated time series forecasting values are predicted in irrigation. The implementation of proposed work is done with the help of banana species in the time period, between Feb 2020, Dec 2020 and 2021. The duration of the cultivation, minimum and maximum water requirements and other supporting parameters are calculated using various sensors. The real and time series water requirements are calculated in the time interval of 12, 24, 36 and 48 etc. With the help of this time series data, month-wise prediction is performed. The implementations were performed in single and group of banana tree water requirements. After implementation of proposed work, for single banana tree 31.4 of the water requirement has been optimized in 2020 time period. Due to this optimized irrigation, heavy usage of fresh water usage and energy wastage is also reduced. © 2022 Elsevier Ltd