Author(s): R. Gonzalez Perea; E. Camacho Poyato; J. A. Rodriguez Diaz
Keywords: Demand modelling and management; Decision-making under uncertainty; Farmer’s behaviour modelling; Artificial intelligence; Irrigation scheduling
Abstract: Irrigated agriculture uses nearly 70 % of the total water consumption in the world. This value accounts for up to 90 % of the total water resources in arid developing countries . In Europe, irrigated agriculture uses around 33 % of total water used although this figure may reach over 80 % in Southern Europe countries  and in the whole Mediterranean region. Thus, efficient water use is essential in a sustainable agricultural system because of reduced water availability mainly in arid and semi-arid regions like Spain. Consequently, water scarcity and the increase in energy demand and their associated costs in pressurized irrigation systems are causing serious challenges for irrigated agriculture and water user associations (WUA). In addition, most of these pressurized irrigation systems has been designed to be operated ondemand where irrigation water is continuously available to farmers complexing the daily decision-making process of the water user association’ managers. Know in advance how much water will be applied by each farmer and its distribution during the day would facilitate the management of the system and would help to optimize the water use and energy costs. Artificial intelligence (AI) techniques have been applied to solve different problems of water resources management and planning. Research also works focused on the prediction of water demand at irrigation district level, using neurogenetic algorithms. However, an optimal daily decision making in WUAs requires not only the knowledge of the of irrigation events occurrence and the applied irrigation depths but their hourly distribution. Since electricity tariffs are organized in hourly periods, the knowledge of the hourly distribution of water consumption would be useful for the optimal contracting of these tariffs. In this work, a hybrid model combining Multiple-input Multiple output (MIMO) of an Adaptive Neuro Fuzzy Inference System (ANFIS) known as Co-active Neuro-Fuzzy Inference System (CANFIS) has been developed to forecast one-day ahead the distribution in energy tariff periods of the irrigation depths applied at farm level. CANFIS has been optimized by the multiobjetive GA NSGA-II . The model which was developed in Matlab™ and integrated as toolbox was validated and tested in a real WUA.