Proposta de um sistema híbrido de produção de energia para veículos aéreos não tripulados
Carregando...
Data
2020-02-12
Autores
Freitas Júnior, Joacy de Lima
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Uma das grandes limitações hoje existente no uso de sistemas aéreos não tripulados com propulsão elétrica, é a pequena autonomia dos voos, o que estimula a pesquisa e desenvolvimento de meios alternativos que proporcionem uma forma de contornar o problema.Esta tese trata do projeto de um Sistema Híbrido de Produção de Energia, onde o controle desse sistema foi implementado com uma lógica PID e uma outra baseada em redes neurais artificiais. A necessidade imposta por alguns tipos de operações com RPA (Remotely-Piloted Aircraft), que exigem uma autonomia mínima, ou ainda que o RPA tenha um sistema alternativo dissimilar de produção de energia para atingir o nível exigido em termos de segurança de voo, são as motivações deste trabalho. Como requisito básico, foi feito uso de materiais e componentes COTS (Commercial off-the-shelf) a fim de viabilizar economicamente a proposta. Para testar o conceito, o sistema foi simulado e construído um protótipo em bancada, sendo que para os testes em voo será utilizado uma aeronave de decolagem e pouso vertical - VTOL (Vertical Take-Off and Landing) que está em fase de construção. Como parte dos trabalhos para se estudar a sua viabilidade, foram feitas simulações em MATLAB®Simulink. Esse método permite que se determine as características físicas e a eficiência do projeto antes que o sistema seja montado. A arquitetura é composta por um motor de combustão interna acionando um gerador para se obter a energia extra necessária para a expansão da autonomia. Apesar deste modelo não ser inédito e já ter sido estudado e testado em projetos anteriores alguns aspectos acrescentaram novos avanços, como neste caso a instalação de um controle inteligente baseado em dados treinados de uma rede neural, visando melhor rendimento e facilidade de alteração para aplicações futuras. O trabalho apresenta resultados positivos mostrando a factibilidade da proposta e abre caminho para o desenvolvimento de um controle mais sofisticado, com a inclusão do algoritmo de treinamento no próprio código, proporcionando o ajuste dos ganhos em tempo real.
Today, one of the great limitations on the use of RPA (Remotely-Piloted Aircraft), is the small autonomy of the flights, which stimulates the research and development of alternative means that provide a way to get around the problem. This thesis deals with the design of a Hybrid Power Generation System, where the control was implemented with a PID logic and another control was based on artificial neural networks. The motivations of this work are the need imposed by the RPA's type of operation for a minimum autonomy, or even if the RPA has an alternative system of energy to reach a level required in terms of flight safety. As a basic requirement, the use of COTS (Commercial off-the-shelf) materials and components was sought to make the proposal economically viable. To test the concept, the system was simulated and tested in bench, and the next step will be the installation in the RPA type VTOL that is being developed. As part of the work to study its viability, simulations were made in MATLAB®Simulink. This method allows to determine the physical characteristics and the efficiency of the design before the system be built. The architecture is composed of an internal combustion engine driving a generator to obtain the extra energy needed to extend the system autonomy. Although this model is not unprecedented and has been studied and tested in previous projects, some aspects will add new advances, as in this case the installation of an intelligent control based on trained data of a neural network, aiming at better performance and ease of change for future developments. The work presents positive results showing the feasibility of the proposal and opens the way for the development of a more sophisticated control, with the inclusion of the training algorithm in the code itself, providing the adjustment of gains in real time.
Today, one of the great limitations on the use of RPA (Remotely-Piloted Aircraft), is the small autonomy of the flights, which stimulates the research and development of alternative means that provide a way to get around the problem. This thesis deals with the design of a Hybrid Power Generation System, where the control was implemented with a PID logic and another control was based on artificial neural networks. The motivations of this work are the need imposed by the RPA's type of operation for a minimum autonomy, or even if the RPA has an alternative system of energy to reach a level required in terms of flight safety. As a basic requirement, the use of COTS (Commercial off-the-shelf) materials and components was sought to make the proposal economically viable. To test the concept, the system was simulated and tested in bench, and the next step will be the installation in the RPA type VTOL that is being developed. As part of the work to study its viability, simulations were made in MATLAB®Simulink. This method allows to determine the physical characteristics and the efficiency of the design before the system be built. The architecture is composed of an internal combustion engine driving a generator to obtain the extra energy needed to extend the system autonomy. Although this model is not unprecedented and has been studied and tested in previous projects, some aspects will add new advances, as in this case the installation of an intelligent control based on trained data of a neural network, aiming at better performance and ease of change for future developments. The work presents positive results showing the feasibility of the proposal and opens the way for the development of a more sophisticated control, with the inclusion of the training algorithm in the code itself, providing the adjustment of gains in real time.
Descrição
Palavras-chave
Inteligência artificial, RPA, VTOL, Artificial intelligence, Energia elétrica - Produção, Aeronaves, Artificial intelligence, Hybrid energy system