Machine Learning Based Load Forecasting
Master thesis
Permanent lenke
https://hdl.handle.net/11250/3017058Utgivelsesdato
2022Metadata
Vis full innførselSamlinger
- Studentoppgaver (TN-IDE) [823]
Beskrivelse
Full text not available
Sammendrag
Population is increasing rapidly and all the demands like electricity are also increasing. The government in England installed smart meters in order to analyze and follow better the energy consumption. Machine learning is the study that gives computers the capability to learn without being explicitly programmed. This is an essential method to transform big data into trends and predictions and forecast energy consumption.In this study, five different machine learning models were executed to predict the energy load consumption. The classical; Linear regression, Random Forest Regression and Multi-layer Perceptron Regression. LSTM (Long Short- Term Memory ) which belongs to artificial neural network and XGBoost or Extreme Gradient Boosting. The goal of this thesis is to forecast the next step short term load energy consumption of 10 houses in London using the machine learning and evaluate the performance of each machine learning model that are implemented.
Based on the result, Linear regression or multiple linear regression outperforms random forest, MLP and XGBoost with an error rate very close to zero on predicting the daily energy consumption. LSTM on the other hand shows a very good result in forecasting the next half hour of the data.