• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.55.2023.tde-14092023-135240
Document
Author
Full name
Yuri Belentani
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Carlos, 2023
Supervisor
Committee
Leandro, Roseli Aparecida (President)
Andrade Filho, Marinho Gomes de
Júnior, José Carlos de Melo Vieira
Zocchi, Silvio Sandoval
Title in Portuguese
Previsão do Preço de Energia no Mercado de Curto Prazo: Uma análise Combinada de Séries Temporais e Redes Neurais Artificiais
Keywords in Portuguese
Ambiente de contratação livre
ARIMA
Inteligência artificial
LSTM
Mercado de energia elétrica brasileiro
Modelagem combinada
Preço da liquidação das diferenças
Redes neurais artificias
SARIMA
Séries temporais
Abstract in Portuguese
Atualmente o Brasil e o mundo passam por uma transição energética, cuja economia de baixo carbono tem se efetivado através de fontes de energia cada vez mais limpas e com operações industriais mais eficientes. Além dos ganhos para o meio ambiente, esse caminho deve trazer benefícios econômicos para os agentes do sistema elétrico brasileiro através das diversas oportunidades que serão geradas. Neste contexto a possibilidade de prever os preços de energia elétrica no mercado de curto prazo (MCP), ambiente de contratação livre, pode contribuir significativamente para a otimização do planejamento de contratação da demanda e uma melhor avaliação dos riscos, seja para os agentes geradores ou para a indústria e grandes consumidores de modo geral. Dessa maneira, buscando contribuir com o aumento de previsibilidade dos agentes que atuam no mercado livre de energia elétrica brasileiro, o presente estudo se propõe prever o valor do Preço de Liquidação das Diferenças (PLD) para quatro semanas operativas subsequentes. Para isso, a ideia é utilizar uma combinação entre modelos AutoRegressivos Integrados de Médias Móveis com Sazonalidade (SARIMA) e redes neurais artificiais recorrentes (LSTM). Primeiramente, serão previstos as quatro semanas operativas subsequentes para as variáveis preditoras e então alimentar uma Rede Neural Artificial (RNA) para produzir o PLD. A metodologia traz como resultado previsões com níveis de acuracidade satisfatórios, principalmente pelo fator combinatório de cada técnica.
Title in English
Energy Price Forecast in the Short Term Market - A Combined Analysis of Time Series and Artificial Neural Networks
Keywords in English
ARIMA
Artificial intelligence
Artificial neural networks
Combined modeling
Free contracting environment
LSTM
Market for Brazilian electricity
Price for settlement of differences
SARIMA
Time series
Abstract in English
Currently, Brazil and the world are going through an energy transition, whose low-carbon economy has been effected through increasingly clean energy sources and with more efficient industrial operations. In addition to the gains for the environment, this path should bring economic benefits to the agents of the Brazilian electrical system through the various opportunities that will be generated. In this context, the possibility of forecasting electricity prices in the short-term market (MCP), a free contracting environment, can contribute to the optimization of demand contracting planning and a greater assessment of risks, whether for generating agents or for industry and large consumers in general. In this way, seeking to contribute to the increase of predictability of the agents that act in the Brazilian electric energy free market, the present study proposes to forecast the value of the difference settlement price (acronym in portuguese PLD) for four subsequent operative weeks. For this, the idea is to use a combination between time series models (SARIMA) and recurrent artificial neural networks (LSTM). Firstly, the four subsequent operative weeks forecast for the predictor variables and then feed the artificial neural network (ANN) to produce the PLD. The methodology results in prediction with strong levels of accuracy, mainly due to the combinatorial factor of each technique.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2023-09-14
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.