Optimization model for irrigation water management in the Dry Zone of Sri Lanka

Abstract

Sri Lanka has a rich irrigation system since ancient times, and the history of irrigation dates back over two millennia and consists of more than thirty thousand lakes. This irrigation system is mainly spread in the Dry Zone in Sri Lanka and mostly using for rice cultivation. In 2019, paddy rice yield for Sri Lanka was 47,954 hg per ha. Paddy rice yield of Sri Lanka increased from 22,485 hg per ha in 1970 to 47,954 hg per ha in 2019 growing at an average annual rate of 1.89%. Irrigation demand for the cultivation is increasing with rising food production. Hence, efficient water management is a key requirement to satisfy the increasing irrigation demand. The current irrigation plan has failed to manage the limited water resources to satisfy the demand. Recent reports on water management in the Dry Zone indicates the water distribution in not meeting the demands of agriculture in terms of reliability, adequacy & timeliness. Reduced wastage of water is a key factor to improving the efficiency of irrigation water management. The main causes of irrigation water wastage are releasing water in excess of the actual irrigation demand and overestimating irrigation demand. The main reason for over estimating is poor consideration of environment factors effects to the crop water requirement. It is possible to reduce the water issue by calculating the most accurate crop water requirement at a particular stage and making an adjustment to the plan while considering environmental factors. The complexity of the irrigation system is the main reason for inaccurate estimation in irrigation issue planning. In computer science, artificial intelligence is the best solution to resolve complex problems. This study used Genetic Algorithms, which fall under Artificial Intelligent to build an optimization model for irrigation water management. This study developed an optimization model for the Rajanganaya Irrigation System and used field data in the 2015 Yala and 2016 Maha seasons to evaluate the final results of the optimization model. Rajanganaya Irrigation System is located in Anuradhapura district of North Central Province which covers an approximate 2500 Ha area with 39 Km canal network. In this study, represented the irrigation system in an appropriate way to use the Genetic Algorithm to recommend an optimum irrigation distribution plan. The model was run for various values of population, generations, cross-over, and mutation probabilities. It is found that the appropriate parameters for population size, number of generations, crossover rate and mutation-rate. Furthermore, the cross-over phase in the Genetic Algorithm operation was modified to get the most accurate result and to apply system limitations such as minimum and maximum irrigation supply.