Forecasting Carbon Intensity and Solar Generation in the Building Sector

Priesol, Matej (2024) Forecasting Carbon Intensity and Solar Generation in the Building Sector. Bachelor's Thesis, Artificial Intelligence.

Preview	Text bAI2024PriesolM.pdf Download (2MB) | Preview
	Text Toestemming Priesol.pdf Restricted to Staff only Download (131kB)

Abstract

The building sector is one of the biggest contributors to climate change and greenhouse gas emissions. However, there are numerous ways in which it can provide benefits to the electrical grid, such as using solar photovoltaic generation for self-sufficient energy production. This study investigates the effectiveness of different forecasting neural network models in order to optimize carbon emissions and solar energy generation. Three different multi-input multi-output models are considered: Multilayer Perceptron, Long Short-Term Memory, and Gated Recurrent Unit. Each model is evaluated using four different criteria: RMSE score for carbon intensity forecasts, RMSE score for solar generation forecasts, training time, and prediction time. The results indicate minimal differences between the models, but the Gated Recurrent Unit achieves slightly lower RMSE scores. However, Multilayer Perceptron is less complex and the training time is a lot faster, with only a small increase in RMSE scores. Therefore, both models are viable options for forecasting carbon intensity and solar generation.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Cardenas Cartagena, J. D. and Fernandes Cunha, R.
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	08 Aug 2024 13:02
Last Modified:	08 Aug 2024 13:02
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/33903

Actions (login required)

View Item