Advanced Machine Learning Techniques for Irrigation System

Abstract

The Water Energy Food (WEF) nexus is an interdependent approach which provides mutual integration for a sustainable ecosystem. WEF ecosystem nexuses provide the solution to achieve long-term environmental, economic and social goals. Particularly, an effective irrigation system optimizes the usage of water, reduces the consumption of energy and increase the food production. An advanced water management system required for agriculture because huge amount of fresh water is wasted while doing the irrigation. An irrigation system is the process of managing water for helping the plants to gets efficient way. The different types of irrigation system used for agriculture such as surface, micro, Drip, and etc. An effective irrigation system required proper requirement analysis. The wireless system, IoT and sensors used to help the requirement and environments analysis. Recent technologies such as IoT, machine and deep learning help to analysis the present and future requirements of water and nutrition of plants. Water and energy consumption in the agriculture industry is not accurately calculated, water efficiency and planning are crucial. An irrigated field waste an enormous quantity of water in irrigation system. The integration of current technology offers a solution for water management and for establishing the right irrigation plan. In this thesis concentrated irrigation requirement analysis using the different parameters for crops. Particularly in this research used banana cultivation for experimental setup and data collections. In this thesis three methodologies used for irrigation requirement analysis and prediction of water requirements. The first methodologies used IoT sensors, reinforcement learning and KNN algorithms for model creations. The second methodology used cloud storage, long short-term memory (LSTM), and adaptive network fuzzy inference system (ANFIS) techniques and transfer learning used for requirement prediction for irrigations. Using the LSTM algorithm short term requirement is analysis, ANFIS is used to calculate the long-term requirement analysis’s and Transfer learning used to reuse the pre-trained model form one farm to other form for better predictions. The third methodology used reinforcement learning KNN algorithm model in the federated learning environments. Using the federated learning better optimized model is created and updated the model using the different parameters. Using this model user data not shared with other farmers. iii The proposed methodologies were implemented using the banana dataset and datasets are collected form location is 8.2473502, 77.2743729,345 (Kanyakumari, Tamil Nadu). The experiments are evaluated using R2, MSLE and accuracy. Using the first methodology 30 to 40% of water is optimized compared to the manual water irrigation. Using the second methodology at 8, 16, 24, 32, and 48 h requirements were predicted. Using the third methodology the accuracy of the predictions was increased from 92.1% to 97.2 %.