A carregar...
Publicação
Preenchimento inteligente de lacunas de dados sensoriais: Aplicação de Machine Learning em sistemas de rega
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 9.97 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Com o crescimento demográfico mundial e as alterações climáticas, a agricultura do século XXI passou
a enfrentar desafios significativos e complexos. O aumento da produtividade tornou-se uma exigência
bem como garantir a sustentabilidade dos recursos disponíveis, pressionando positivamente para
mudanças nos métodos de produção e para a gestão eficiente desses recursos.
A rega é uma das práticas culturais com maior responsabilidade no crescimento vegetativo operando
em paralelo com a sustentabilidade. Produzir mais com menos tornou-se o mote dos agricultores nos
dias de hoje, sendo crucial minimizar o desperdício e assegurar que, neste caso, a rega atenda às
necessidades especificas das culturas em todas as fases do seu ciclo.
O avanço das tecnologias de comunicação e a adoção a equipamentos IoT (sensores, sondas de
humidade e outras ferramentas de monitorização) permitiu ao agricultor uma melhoria significativa nas
tomadas de decisão, promovendo insights futuros com base nos dados recolhidos. Este é um dos
objetivos do trabalho realizado pela empresa HIDROSOPH, que desempenha um papel vital neste
processo, oferecendo soluções inovadores na otimização e gestão eficiente da água.
Assim, em parceria com a HIDROSOPH, foi proposto um trabalho de validação e consolidação dos
dados recolhidos pelos equipamentos IoT em funcionamento, garantindo a uniformidade, integridade e
aplicabilidade dos dados nos diversos serviços prestados pela empresa. O caso de estudo concentrouse
em dados de 12 explorações de olival em sebe, instalado em várias regiões de Portugal Continental,
respetivo ao ano 2023. Verificou-se um problema recorrente em quase todas as explorações, que
consiste na interrupção no registo de dados devido a falhas técnicas dos sensores.
Para abordar este problema, foi utilizado o modelo XGBoost de Machine Learning, aplicado para
imputar valores previstos nas lacunas de dados. Apesar da execução prática ter sido focada no
XGBoost, outros modelos foram avaliados teoricamente, tendo-se concluído que o XGBoost é a
solução mais adequada para garantir a integridade contínua dos dados e suportar os algoritmos
utilizados nos serviços da empresa.
A resolução dessas falhas é essencial para garantir a eficácia e a fiabilidade das soluções providas
pela HIDROSOPH.
With global population growth and climate change, agriculture in the 21st century is facing significant and complex challenges. Increasing productivity has become a requirement, as has ensuring the sustainability of available resources, putting positive pressure on changes in production methods and the efficient management of these resources. Irrigation is one of the cultural practices with the greatest responsibility for vegetative growth, operating in parallel with sustainability. Producing more with less has become the motto for farmers these days, and it is crucial to minimize waste and ensure that, in this case, irrigation meets the specific needs of crops at all stages of their cycle. Advances in communication technologies and the adoption of IoT equipment (sensors, humidity probes and other monitoring tools) have enabled farmers to significantly improve their decision-making, promoting future insights based on the data collected. This is one of the objectives of the work carried out by the company HIDROSOPH, which plays a vital role in this process, offering innovative solutions in the optimization and efficient management of water. So, in partnership with HIDROSOPH, we proposed a project to validate and consolidate the data collected by the IoT equipment in operation, guaranteeing the uniformity, integrity and applicability of the data in the various services provided by the company. The case study focused on data from 12 highdensity olive grove farms from different locations in mainland Portugal, for the year 2023. There was a recurring problem on almost all of the farms, which consisted of interruptions in data recording due to technical failures in the sensors. To address this problem, the XGBoost Machine Learning model was used to impute predicted values into data gaps. Although the practical implementation was focused on XGBoost, other models were evaluated theoretically, and it was concluded that XGBoost is the most suitable solution for ensuring continuous data integrity and supporting the algorithms used in the company's services. Resolving these gaps is essential to maintain the effectiveness and reliability of the solutions provided by HIDROSOPH.
With global population growth and climate change, agriculture in the 21st century is facing significant and complex challenges. Increasing productivity has become a requirement, as has ensuring the sustainability of available resources, putting positive pressure on changes in production methods and the efficient management of these resources. Irrigation is one of the cultural practices with the greatest responsibility for vegetative growth, operating in parallel with sustainability. Producing more with less has become the motto for farmers these days, and it is crucial to minimize waste and ensure that, in this case, irrigation meets the specific needs of crops at all stages of their cycle. Advances in communication technologies and the adoption of IoT equipment (sensors, humidity probes and other monitoring tools) have enabled farmers to significantly improve their decision-making, promoting future insights based on the data collected. This is one of the objectives of the work carried out by the company HIDROSOPH, which plays a vital role in this process, offering innovative solutions in the optimization and efficient management of water. So, in partnership with HIDROSOPH, we proposed a project to validate and consolidate the data collected by the IoT equipment in operation, guaranteeing the uniformity, integrity and applicability of the data in the various services provided by the company. The case study focused on data from 12 highdensity olive grove farms from different locations in mainland Portugal, for the year 2023. There was a recurring problem on almost all of the farms, which consisted of interruptions in data recording due to technical failures in the sensors. To address this problem, the XGBoost Machine Learning model was used to impute predicted values into data gaps. Although the practical implementation was focused on XGBoost, other models were evaluated theoretically, and it was concluded that XGBoost is the most suitable solution for ensuring continuous data integrity and supporting the algorithms used in the company's services. Resolving these gaps is essential to maintain the effectiveness and reliability of the solutions provided by HIDROSOPH.
Descrição
Mestrado em Ciência de Dados em Agricultura, Alimentação, Floresta e Ambiente/ Instituto Superior de Agronomia, Universidade de Lisboa
Palavras-chave
rega sensores IoT machine learning sustentabilidade XGBoost irrigation IoT sensors machine learning sustainability
Contexto Educativo
Citação
Pinto, D.E. Preenchimento inteligente de lacunas de dados sensoriais: Aplicação de Machine Learning em sistemas de rega. Lisboa: ISA, 2024, 88 p. Dissertação de Mestrado
Editora
Instituto Superior de Agronomia, Universidade de Lisboa