Trabalho de Conclusão de Curso de Especialização
Modelos de previsão na otimização de usinas elétricas virtuais
Fecha
2022-07-20Autor
Mansilha, Marcio Burger
Institución
Resumen
The work analyzed ARIMA models for forecasting day-ahead price clearing markets. The
models were based on time series analysis and provide reliable and accurate market price
predictions. The objective was to verify the quality of forecasting electricity prices, which are
unavoidable aspects of a VPP. The methodology used consisted of a bibliographic review of
the VPP's and the method proposed by Box-Jenkins, which is also known as ARIMA
methodology, was used to perform forecasting in time series. As a method, a structured case
study was carried out to simulate ARIMA forecast models with real data collected from the spot
prices of the next day in Great Britain in GBP. The data covered the period from 2015 to
September 30, 2020.The prediction result with neural networks presented the best performance
measures for the ARIMA NNETAR model. It is concluded that the use of the time series
methodology constitutes an important support and support in the PPV forecasts as it allows for
better decision making.