Proposta de um modelo preditivo de resultados de leilões de transmissão de energia elétrica

Abstract

The dissertation presents a proposal for a predictive model based on Bayesian linear regression to analyze and predict the results of the difference between the Permitted Annual Revenue defined in the notice and the Permitted Annual Revenue proposed in electricity transmission auctions in Brazil. In a scenario of accelerated growth in energy generation and transmission system, transmission auctions play an important role in the expansion and modernization of the national electrical infrastructure. The study highlights the strategic importance of these auctions as a fundamental mechanism for the efficient allocation of resources, ensuring the expansion of the transmission network in an economic and sustainable way. For this study, a multiple linear regression model was initially used, where it was found that the variables used in this model explained approximately 46.8% of the variability of the dependent variable. After obtaining unsatisfactory data, he switched to a Bayesian linear regression technique. When analyzing historical data from transmission auctions, the Bayeasian model was able to explain approximately 87% of the variability in the bid value, which means a high value when compared to other studies and knowing that most of the variables raised are contained in the auction notices. It is important to highlight that the most relevant variables identified by the model were the variables number of projects, length of the transmission line, projects carried out in the northeast subsystem, deadline for executing the project and number of competitors, indicating their strong influence on the results of the auctions. This finding offers insights to market participants, allowing them to better understand which factors may impact their bidding strategies. The relevance of this study is not limited to its practical application. It highlights the crucial need for an in-depth understanding of transmission auctions in the context of Brazilian energy development. By studying these events, not just as economic transactions, but as sensitive indicators of market dynamics, a more efficient and competitive energy sector in Brazil can be promoted. This dissertation not only proposes a predictive model using Bayesian linear regression for electricity transmission auctions in Brazil, but also highlights the strategic importance of understanding these auctions, contributing to broader knowledge about the mechanisms of the electricity market.