Day-ahead electricity price forecasting for Türkiye using an ensemble machine learning technique

Date

2024-11

Author

Özbudak, Çağkan

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

102
views

13
downloads

Cite This

In a liberal electricity market where there is competition, accurate hourly electricity price forecasting is important. Electricity producers and consumers require methods for precise price predictions. Producers and consumers may organize their bidding strategies to maximize their benefits by using price projections, which provide important information. Due to the under-maturation and low proliferation of grid-scale storage technologies, the increasing uncertainty with the high penetration of intermittent technologies such as solar and wind makes forecasting more challenging and critical than ever before. Therefore, changes in supply or demand occur with an impact on pricing. Moreover, economic instability mainly originated from national monetary policies together with the political conjoncture in the neighbouring countries, which are also energy suppliers, in the recent decade decrease the predictability of the prices. In this thesis, XGBoost, SVR and an ensemble of these two algorithms are used for precise for precise and reliable day-ahead electricity price forecasting in the electricity market in Türkiye. The proposed algorithms are compared with other benchmark models which are which are SARIMA and Naive Models for precise and reliable day-ahead electricity price forecasting in the electricity market in Türkiye. Different model settings and time periods for the performance metrics are investigated. The results obtained indicate that the proposed method used is promising in terms of performance metrics which shows competing values compared to the benchmark models and other studies in the literature.

Subject Keywords

Day-ahead electricity price, Price forecasting, Machine learning, Ensemble learning, XGBoost, Support vector regression (SVR), SARIMA, Naive models

URI

https://hdl.handle.net/11511/112648

Collections

Graduate School of Natural and Applied Sciences, Thesis