In this paper day-ahead electricity price forecasting for the Denmark-West region is realized with a 24 h forecasting range. The forecasting is done for 212 days from the beginning of 2017 and past data from 2016 is used. For forecasting, Autoregressive Integrated Moving Average (ARIMA), Trigonometric Seasonal Box-Cox Transformation with ARMA residuals Trend and Seasonal Components (TBATS) and Artificial Neural Networks (ANN) methods are used and seasonal naïve forecast is utilized as a benchmark. Mean absolute error (MAE) and root mean squared error (RMSE) are used as accuracy criterions. ARIMA and ANN are utilized with external variables and variable analysis is realized in order to improve forecasting results. As a result of variable analysis, it was observed that excluding temperature from external variables helped improve forecasting results. In terms of mean error ARIMA yielded the best results while ANN had the lowest minimum error and standard deviation. TBATS performed better than ANN in terms of mean error. To further improve forecasting accuracy, the three forecasts were combined using simple averaging and ANN methods and they were both found to be beneficial, with simple averaging having better accuracy. Overall, this paper demonstrates a solid forecasting methodology, while showing actual forecasting results and improvements for different forecasting methods.
This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited