DIGITAL LIBRARY
PRACTICAL SESSION FOR THE SUBJECT ‘IRRIGATION AND DRAINAGE’. ARTIFICIAL NEURAL NETWORKS IN IRRIGATION ENGINEERING: A VALUABLE ALTERNATIVE TO CONVENTIONAL MODELING APPROACHES
Universitat Politecnica de Valencia (SPAIN)
About this paper:
Appears in: ICERI2013 Proceedings
Publication year: 2013
Pages: 4520-4528
ISBN: 978-84-616-3847-5
ISSN: 2340-1095
Conference name: 6th International Conference of Education, Research and Innovation
Dates: 18-20 November, 2013
Location: Seville, Spain
Abstract:
One of the most important professional competences of Agricultural Engineers is the ability to perform design and management of irrigation installations. Therefore several subjects on the fundamentals of irrigation and drainage engineering are scheduled in the degree programs of Agricultural Engineering.

Artificial Neural Networks (ANNs) are massively parallel distributed processors consisting of simple processing units, which have a natural propensity for storing experimental knowledge and making it available for use. They have shown to be efficient and less-consuming in the modeling of complex systems.

In the last 10 years ANNs have hold the attention of many researchers providing very satisfactory results in an increasing number of agricultural applications, including irrigation. Hence, ANNs appear as a very robust alternative to conventional approaches for estimating irrigation targets.
This paper proposes a new practical session for the subject ‘Irrigation and Drainage’ corresponding to the degree of Agricultural Engineering at the Universitat Politècnica de València. This practical session aims at introducing to the students these new modeling tools and their associated abilities for estimating target variables in irrigation in comparison to conventional procedures.
The session is scheduled as follows. First, the fundamentals of ANNs are briefly introduced. In a second part, four specific cases corresponding to four important target variables are studied using the software Matlab. The lecturer provides the files containing the required Matlab code, and the students perform on-site those ANN simulations.

In the first case, students assess the accuracy of conventional and neural models for estimating reference evapotranspiration, the key variable for the definition of crop water requirements. Different climatic scenarios and input combinations are discussed. In the second case, local head losses are estimated according to both approaches. Accurate head loss prediction is necessary for a suitable hydraulic design of irrigation installations. The third target variable is stem water potential, an optimum indicator of plant water status, usually used in irrigation scheduling. Given that its experimental measurement is very time consuming and cannot be automated, its estimation turns into a task of great relevance. Finally, outlet dissolved oxygen in sand filters is a key parameter when the installation is fed with effluents and a suitable filtration is mandatory.
Students get in touch with a new and valuable modeling tool in this field, as well as with current research studies in this area. They learn how to extract and apply knowledge from research papers and put it into practice. This session also contributes to visualize the applicability and usefulness of theoretical contents of the subject, improving motivation and learning. This contextualization of contents also allows the students to discern their skills and competencies as engineers and/or researchers. The lecturer also makes known a part of his research field allowing future collaborations with interested students.
Keywords:
Practical session, artificial neural networks, matlab, modeling, irrigation.