University of Twente Student Theses
Comparative study of the state of the art of machine learning-based models for load forecasting
Abdul Wahab, Aiman (2024) Comparative study of the state of the art of machine learning-based models for load forecasting.
This is the latest version of this item.
|
PDF
658kB |
| Abstract: | A reliable energy forecasting system is crucial in the energy sector. Multiple classical machine learning models have been investigated in terms of accuracy. This paper aims to shed light on two hybrid models: CNN-XGBoost and LSTM-XGBoost for short-term load forecasting (STLF). The Convolutional Neural Network (CNN) and Long Short-term Memory (LSTM) models can be used to extract relevant time series patterns which can then be passed on to XGBoost and perform better forecasting. The model is evaluated on the load energy data collected from a company residing in a business park in the Netherlands. Temperature and weather data were also considered. The two gradient boosting models were compared, and XGBoost was chosen for constructing a hybrid model because it performed better than CatBoost. The hybrid models were compared and performed better than XGBoost in most benchmarks. Between the two hybrid models, CNN-XGBoost outperformed LSTM-XGBoost by a small margin. Additionally, the paper explores an interpretability tool, Maximum Mean Discrepancy (MMD-Critic) with limited research in the time series domain and is used for understanding black-model decision-making. The model performed worse when attempting to forecast the criticism date with a MAE = 7.23 compared to the preceding two and three weeks with scores of 5.85 and 6.01. The paper demonstrated that hybrid models are more reliable than decision tree models and additionally presented a tool for understanding the weak links of a hybrid model. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Computer Science BSc (56964) |
| Link to this item: | https://purl.utwente.nl/essays/100778 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page