Neural network models to predict the electricity and final energy demand of Portugal

23 de julho de 2020

Energy is one of the most important resources for the growth and development of a country in the modern society, and reliable energy forecasting is essential for the effective management of this resource. In this thesis, an artificial neural network technique was applied to the long term forecasting of the electricity and final energy needs of Portugal for the period of 2015 – 2050, using scenarios of socioeconomic variables as network inputs. The artificial neural network is a computing system that learns by acquiring experience, like the human brain does, and the type of network used here was the recurrent neural network, a model sensible to the temporal sequence of the data. This method yielded a relative error lower than 3% for the forecasting of both resources, which makes it two times better than the linear regression, a commonly used approach in this type of problem. Additionally, these forecasts were compared to the ones obtained by the Portuguese government. Regarding the final energy consumption, the predictions obtained suggest that the consumption values should be at least 1.5 times higher than the ones obtained by the government. On the other hand, the algorithm cannot accurately predict the electricity consumption if the government follows through with the sustainable measures proposed. Despite this, it was considered that the electricity consumption values estimated by the government are optimistic, even though they might be viable.

Autores da comunidade :

• 

  J

  João Pedro Pascoal Martins Sanchez Gonzalez

  ist178649

Orientadores desta instituição:

• 

  M

  Miguel Afonso Dias de Ayala Botto

  ist12889

• 

  T

  Tânia Alexandra dos Santos Costa e Sousa

  ist14067

mechanical-engineering - Engenharia Mecânica

por - Português

Instituto Superior Técnico