(149n) Machine Learning-Based Predictive Irrigation Scheduling

Water scarcity and climate change are placing increasing pressure on the agricultural sector to conserve water resources while optimizing crop yield [1]. Traditional irrigation scheduling methods have limitations in terms of efficiency and practicality for large-scale irrigation systems [2]. Machine learning techniques and predictive control have emerged as promising solutions to optimizing irrigation scheduling [3].

This work proposes a machine-learning-based predictive irrigation scheduler that employs the three paradigms of machine learning to improve water use efficiency and sustainability in agriculture: supervised learning, unsupervised learning, and reinforcement learning. Specifically, the k-means clustering method will be used in the unsupervised learning paradigm to delineate the considered field into definite management zones based on soil hydraulic parameters and topology information. Next, a long short-term memory network will be employed in the supervised learning paradigm to train dynamic models for the resulting management zones. These dynamic models will be used in the design of a mixed-integer model predictive control, where the irrigation scheduler's main objective is to determine the daily irrigation decision and irrigation rate that ensures optimal root water uptake while reducing water wastage and energy consumption during the irrigation event.

Given that mixed-integer problems are challenging to solve, the reinforcement learning paradigm will be used to determine a policy that calculates the daily irrigation decision. This decision will be communicated to the model predictive controller, which can then determine the daily irrigation rate.

Finally, the proposed scheduler will be evaluated under dry and wet weather conditions. The results show that it can achieve a 15% and 22% savings in water under dry and wet weather conditions, respectively.

References

[1]. United Nations World Water Assessment Programme. The United Nations World Water Development Report 2018: Nature-Based Solutions for Water, 2018

[2]. Shah, Sirish L., et al. "Meeting the challenge of water sustainability: The role of process systems engineering." AIChE Journal 67.2 (2021): e17113.

[3]. Saggi, Mandeep Kaur, and Sushma Jain. "A Survey Towards Decision Support System on Smart Irrigation Scheduling Using Machine Learning approaches." Archives of Computational Methods in Engineering 29.6 (2022): 4455-4478.