Machine Learning Approach for Trend Prediction to Improve Returns on Brazilian Energy Market

Author(s):Show less -

Santos, Moises R. ; Braz, Douglas D. C. ; Carvalho, Andre C. P. L. F. ; Tinos, Renato ; Paula, Marcos B. S. ; Doretto, Gabriel ; Guarnier, Ewerton ; Filho, Donato Silva ; Suiama, Danilo Y. ; Ferreira, Lorena E. ; Carmo, Jose E., Jr. ; IEEE

• Document type: Journal article
• Source: 2022 IEEE LATIN AMERICAN CONFERENCE ON COMPUTATIONAL INTELLIGENCE (LA-CCI); v. N/A, p. 6-pg., 2022-01-01.

Abstract

The Free Energy Market in Brazil has changed a lot since its implementation in the late 1990s. Its expansion has accelerated in recent years, making operations more complex, dynamic, and liquid. In addition, implementing an over-thecounter Energy Exchange (BBCE), an environment where market agents can purchase and sell energy for trading or deliver it to their customers, resulted in a good set of historical data. Although most works focus on predicting the price trend of the financial market, it is possible to apply a part of these tools in the electricity markets, considering their particularities. This study aims to present a new methodology for the day ahead trend prediction in the Brazilian energy market using machine learning-based strategies. We propose input time lags, technical measures, and seasonal features for trend prediction and a new approach for combining predictions based on a Mixture of Experts. The results show that the proposed combination of predictions strategy results better than the others in cumulative return. For price trend prediction in the Brazilian energy market, the system based on a Mixture of Experts for combining predictions with optimization of different metrics shows promise for improving financial metrics. (AU)

• FAPESP's process: 13/07375-0 - CeMEAI - Center for Mathematical Sciences Applied to Industry
• Grantee: Francisco Louzada Neto
• Support Opportunities: Research Grants - Research, Innovation and Dissemination Centers - RIDC
• FAPESP's process: 19/10012-2 - Meta-learning for time-series forecasting
• Grantee: Moisés Rocha dos Santos
• Support Opportunities: Scholarships in Brazil - Doctorate