Heating energy consumption forecasting based on machine learning
Trotskii, Igor (2018)
Trotskii, Igor
Hämeen ammattikorkeakoulu
2018
All rights reserved
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2018060412296
https://urn.fi/URN:NBN:fi:amk-2018060412296
Tiivistelmä
The author’s aim in this thesis project was to develop a machine learning model, which could create short-term forecasts regarding heating energy consumption of a building. Even short-term energy consumption forecasts can have a major impact on building automation and energy distribution systems. Possible application spheres include smart grid development and simpler maintenance.
A feed forward artificial neural network was designed as a result of examination and testing of different models in order to get the most accurate predictions possible. To create an effective neural network various loss and activation functions as well as optimizers were reviewed.
To obtain better results some preprocessing techniques were applied to filter corrupted and unreliable data. The designed model was successfully trained to perform forecasting on data from the same distribution as the training data.
A feed forward artificial neural network was designed as a result of examination and testing of different models in order to get the most accurate predictions possible. To create an effective neural network various loss and activation functions as well as optimizers were reviewed.
To obtain better results some preprocessing techniques were applied to filter corrupted and unreliable data. The designed model was successfully trained to perform forecasting on data from the same distribution as the training data.