Tesis y Tesistas 2020 - Postgrado - Fac. de Informática - UNLP

MAESTRÍA
INGENIERÍA DE SOFTWARE
Mg. Ernesto Esteban Ledesma
e-mail
erneledesma@gmail.com
Advisor
Dra. Daniela López de Luise
Codirector
Mg. Javier Bazzoco
Thesis defense date
November 20, 2020
SEDICI
http://sedici.unlp.edu.ar/handle/10915/110794
Prediction model of
risk in water resources
for precision agriculture
Keywords: Temporary mining, Diffuse Harmonic Systems, Diffuse Patterns, Precision Crops, Rice Crops.
Motivation
In recent years, precision agriculture has become very
important in the search to optimize resources and obtain
better results, day by day we work on new technologies that
allow reaching the producer to obtain these improvements.
In crops that require a high demand for water, such as rice,
different irrigation techniques are applied to lower the
water demand and continue to obtain high yields. However,
there are still difficulties to perform optimal monitoring in
real time.
This work is developed in the IDTILAB of the Faculty of
Science and Technology of UADER (Concepción del Uruguay,
Entre Ríos), in conjunction with the section of INTA
(National Institute of Agricultural Technology, Concepción
del Uruguay), and presents a model of Innovative behavior
and prototype to monitor precision crops in real time.
Founded on the latest in temporal data mining, it uses an
extension of the well-known Harmonic Systems (HS) (Lopez
de Luise D. 2013) called Fuzzy Harmonic Systems (FHS) (Lopez
de Luise D. 2013a, 2013b) (Bel W. 2018) that constitutes a
simple and lightweight heuristic capable of detecting and
predicting critical water stress events in rice crop lots. The
functional prototype of KRONOS.AgroData and KRONOS.
AgroMonitor that implements the adapted FHS model for
predicting the level of drought risk in the irrigation plots in
rice crops in the San Salvador area located in (Entre Ríos)
and in the INTA area, Concepción del Uruguay (Entre Ríos).
This prototype is made with Arduino technology for data
acquisition and web technologies such as React®, NextJS®,
NodeJS® and MQTT®.
The design allows evaluating the performance and efficiency
of the proposed model in a real field test environment where
variables of various types are involved (weather, soil moisture
variations, soil water level, PH, among others).
From the field studies and statistical analyzes shown in
this work, it can be affirmed that the derived model allows
determining sampling and irrigation intervals much more
suitable than the traditional ones, and satisfactorily
evaluating the yields and cultivation conditions.
The specific objective proposed for this work consists of the
appropriate validation of the proposed model.
Thesis contributions
In the thesis work, two prototypes were developed for the
analysis of the model with low cost technologies to make
predictions about critical water events in rice crops.
The model implements harmonic systems and fuzzy
harmonic systems. In the findings, it can be said that the
pattern detection method is flexible enough to apply it to
the context of detecting water stress in crops, its simplicity
allows profiling and reduction of computational complexity.
The FHS (Fuzzy Harmonic System) model effectively detects
fuzzy temporal patterns that allow the model to analyze
data from a more flexible perspective.
The FHS model is implemented in the KronosData and
KronosAgro prototypes in order to carry out the technical
validation of its operation in real environments of rice
crops. A priori, the tests carried out indicate that the
operation of the data acquisition and web prototype (in a
real environment of controlled field tests) it is deduced that
the behavior of the model maintains its efficiency even in
a real variable environment that requires processing and
immediate response.
88