Name: | Description: | Size: | Format: | |
---|---|---|---|---|
4.23 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Face ao desafio da transição energética, a energia solar fotovoltaica apresenta-se cada vez mais como
uma solução de geração de energia limpa para o fornecimento de eletricidade tanto a habitações como
empresas. Para que tal seja possível, é necessário tornar esta tecnologia o mais desenvolvida possível
e reduzir os seus custos para que se torne cada vez mais rentável. De modo a responder a este desafio,
foi criada a digital twin, uma ferramenta que permite modular a produção de um sistema fotovoltaico
e comparar a sua simulação com as medições reais. Deste modo a digital twin permite antecipar problemas e reduzir os custos de manutenção dado que apenas é necessário realizar manutenção quando a
ferramenta se depara com uma discrepância entre a simulação da produção fotovoltaica e as medições
reais. Esta dissertação descreve o funcionamento da versão original da digital twin, desenvolvida pela
EDP comercial, a análise da média dos erros aleatório (nRMSE) e sistemático (nMBE) da simulação da
produção de 20 instalações fotovoltaicas e as áreas do programa que necessitam de ser melhoradas. Posteriormente são testados novos modelos de temperatura, de radiação difusa e de perdas estáticas através
da análise dos erros aleatório e sistemático, dando origem à digital twin 2.0, otimizada para o valor médio dos erros das 20 instalações, inferior ao observado na versão original. A aplicação de um modelo de
cálculo da temperatura do módulo fotovoltaico revelou-se determinante para a redução do erro aleatório
médio enquanto que para o modelo de radiação difusa anisotrópico não se observou um efeito relevante.
A versão original da digital twin apresenta um nRMSE médio de 5.2% e um nMBE médio de 4% para
as 20 instalações estudadas, sendo que a implementação da versão 2.0 resultou na redução dos valores
médios de nRMSE e nMBE para 4.9% e 0.03% respetivamente.
Energy transition presents the main challenge of our century, and photovoltaic energy is becoming the solution for a supply of green electricity to homes and businesses. However, it is essential to continue improving photovoltaic technology and reducing its cost. The digital twin was developed with this in mind. It is a program that modulates the photovoltaic production of an installation and compares it to the measurements provided. The digital twin allows to anticipate faults and can help decrease maintenance costs since maintenance is only required when there is a discrepancy between the simulation data and the measurements, meaning there is a fault in the photovoltaic system. This dissertation describes how the digital twin 1.0 operates, analyses the mean value of the root mean square error and mean bias error for 20 installations and identifies the areas of the program that need to be enhanced. Posteriorly new models of temperature, diffuse irradiance and static losses are tested by the mean simulation error analysis for the 20 installations, creating the digital twin 2.0. The modulation of the photovoltaic module temperature represents the main improvement of the mean bias error, although the implementation of the anisotropic diffuse irradiance does not correspond to an improvement in the mean simulation errors. For the 20 installations studied in this dissertation, the original version of the digital twin has a root mean square error of 5.2% and a mean bias error of 4%, with the implementation of version 2.0 the nRMSE decreased to 4.9% while the nMBE reduced to 0.03%.
Energy transition presents the main challenge of our century, and photovoltaic energy is becoming the solution for a supply of green electricity to homes and businesses. However, it is essential to continue improving photovoltaic technology and reducing its cost. The digital twin was developed with this in mind. It is a program that modulates the photovoltaic production of an installation and compares it to the measurements provided. The digital twin allows to anticipate faults and can help decrease maintenance costs since maintenance is only required when there is a discrepancy between the simulation data and the measurements, meaning there is a fault in the photovoltaic system. This dissertation describes how the digital twin 1.0 operates, analyses the mean value of the root mean square error and mean bias error for 20 installations and identifies the areas of the program that need to be enhanced. Posteriorly new models of temperature, diffuse irradiance and static losses are tested by the mean simulation error analysis for the 20 installations, creating the digital twin 2.0. The modulation of the photovoltaic module temperature represents the main improvement of the mean bias error, although the implementation of the anisotropic diffuse irradiance does not correspond to an improvement in the mean simulation errors. For the 20 installations studied in this dissertation, the original version of the digital twin has a root mean square error of 5.2% and a mean bias error of 4%, with the implementation of version 2.0 the nRMSE decreased to 4.9% while the nMBE reduced to 0.03%.
Description
Tese de mestrado, Engenharia da Energia e Ambiente, 2023, Universidade de Lisboa, Faculdade de Ciências
Keywords
Digital twin Simulação Medições Energia Solar Fotovoltaica Teses de mestrado - 2023