Akademik Veri Yönetim Sistemi
12th INTERNATIONAL CONGRESS ON ENGINEERING, ARCHITECTURE AND DESIGN, 23 - 25 Aralık 2023, ss.324-333
The accuracy of electricity price forecasting is critical in ensuring the reliability
and efficiency of power industry operations. Accurate forecasting can be accomplished by
employing advanced artificial intelligence models, such as LSTM and CNN-LSTM, which
are known for their strong analysis capabilities in time series data, allowing the discovery
of hidden patterns. The primary goal of this research is to evaluate the performance of the
CNN-LSTM model, which uses historical electricity and natural gas prices to perform
predictive analysis. This study emphasizes the importance of natural gas prices in the
Turkish electricity market as an exogenous variable when forecasting electricity prices in
a country where natural gas accounts for 22.91% of total energy generation. The EXIST
Transparency Platform provided 5-year historical data, which included hourly records for
electricity prices and daily data for natural gas prices. This historical data is used to
forecast electricity prices seven days in advance. Multiple measures of model accuracy
were used, including mean absolute error, root mean squared error, mean absolute
percentage error, and forecast accuracy in line with the correct trends. The results demonstrated the superiority of the CNN-LSTM model over the LSTM models. When the
exogenous variable was included, the CNN-LSTM model performed the best, with accurately forecasted trends of electricity prices for six out of seven days.