Aplicación de Redes Neuronales en controladores de baterías
Carregando...
Data
2019-12-16
Autores
Sánchez Torres, Norah Nadia
Título da Revista
ISSN da Revista
Título de Volume
Editor
Resumo
El presente estudio tiene como objetivo aplicar el uso de redes neuronales artificiales, que tiene como una de sus aplicaciones operar como un aproximador universal de funciones, mapeando la relación funcional entre las variables de un sistema a partir de un conjunto conocido de valores muestreados. En este contexto, este trabajo aborda un método para predecir el estado de carga de las baterías utilizando técnicas de redes neuronales artificiales a través de una base de datos y modelos de la curva de carga de las baterías de cloruro de sodio y níquel; y así analizar el comportamiento del sistema de gestión de baterías, a través de los modelos encontrados en las curvas de salida. Así este estudio en principio presenta una breve introducción del mercado de energía, seguido de la justificativa y motivación que llevaron a desarrollar el mismo. En seguida, se presenta la metodología empleada, en el software MATLAB, paso a paso para la obtención de las curvas de carga. Seguido de una breve descripciones de los sistemas de almacenamiento de energía, baterías. Continuando con una descripción del sistema de gerenciamiento de baterías, funcionalidades y aplicaciones. Y por fin, una descripción de redes neuronales, clasificación arquitectura y aplicaciones en ingeniería; que, para el caso, el método propuestos utiliza una red neuronal artificial Perceptron multicapa, una arquitectura de avance (feedforward) con algoritmo de entrenamiento de retropropagación. Con todo esto, finalmente, los resultados indican la capacidad del método para indicar el estado de carga de la batería, así como el análisis de los errores estipulados. Concluyendo que, la configuración utilizada tiene un mejor rendimiento al ajustar el número de capas, y puede ser aplicado en otras baterías, como es el caso de la batería de litio; con los errores y percances encontrados a lo largo de este estudio se presenta algunos trabajos a futuro.
The present study aims to apply the use of artificial neural networks, which has as one of its applications to operate as a universal approximator of functions, mapping the functional relationship between the variables of a system from a known set of sampled values. In this context, this work addresses a method to predict the state of charge of batteries using artificialneural network techniques through a database and models of the charge curve of batteries of sodium chloride and nickel; and thus analyze the behavior of the battery management system, through the models found in the output curves. Thus, this study in principle presents a brief introduction to the energy market, followed by the justification and motivation that led to its development. Next, the methodology used is presented, in the MATLAB software, step by step to obtain the load curves. Followed by a briefdescription of the energy storage systems, batteries. Continuing with a description of the battery management system, features and applications. And finally, a description of neural networks, architectural classification and applications in engineering; that, for that matter, the proposed method uses a multilayer Perceptron artificial neural network, a feedforward architecture with backpropagation training algorithm. With all this, finally, the results indicate the ability of the method to indicate the state of charge of the battery, as well as the analysis of the stipulated errors. Concluding that, the configuration used has a better performance when adjusting the number of layers, and can be applied in other batteries, as in the case of the lithium battery; With the errors and mishaps found throughout this study, some future work is presented.
The present study aims to apply the use of artificial neural networks, which has as one of its applications to operate as a universal approximator of functions, mapping the functional relationship between the variables of a system from a known set of sampled values. In this context, this work addresses a method to predict the state of charge of batteries using artificialneural network techniques through a database and models of the charge curve of batteries of sodium chloride and nickel; and thus analyze the behavior of the battery management system, through the models found in the output curves. Thus, this study in principle presents a brief introduction to the energy market, followed by the justification and motivation that led to its development. Next, the methodology used is presented, in the MATLAB software, step by step to obtain the load curves. Followed by a briefdescription of the energy storage systems, batteries. Continuing with a description of the battery management system, features and applications. And finally, a description of neural networks, architectural classification and applications in engineering; that, for that matter, the proposed method uses a multilayer Perceptron artificial neural network, a feedforward architecture with backpropagation training algorithm. With all this, finally, the results indicate the ability of the method to indicate the state of charge of the battery, as well as the analysis of the stipulated errors. Concluding that, the configuration used has a better performance when adjusting the number of layers, and can be applied in other batteries, as in the case of the lithium battery; With the errors and mishaps found throughout this study, some future work is presented.
Abstract
Descrição
Trabajo de final de Curso presentado al Instituto Latino-Americano de Tecnología, Infraestructura y Territorio de la Universidad Federal de Integración Latinoamericana, como requisito para obtener el título de Bachiller en Ingeniería de Energías. Orientador:Dr. Jorge Javier Gimenez Ledesma y Coorientador:Dr. Oswaldo Hideo Ando Junior
Palavras-chave
Estado de carga, Redes Neuronales Artificiales, Software MATLAB
Citação
SÁNCHEZ TORRES, Norah Nadia. Aplicación De Redes Neuronales En Controladores De Baterías. 2019. 60 p. Trabalho de Conclusão de Curso (Graduação em Engenharia de Energias) – Universidade Federal da Integração Latino – Americana, Foz do Iguaçu, 2019.