| dc.description.abstract | The European power markets have become increasingly volatile, and being able to make accurate forecasts in this environment has become more important than ever before. This thesis aims to accurately predict the North Sea link day ahead prices using a selection of autoregressive models. The autoregressive models considered are the autoregressive moving average, neural net autoregressive and wavelet autoregressive moving average plus neural network autoregressive. Actual and forecasted data of the North Sea link prices and several predictors was downloaded during the period 1 October 2021 to 31 December 2022. The data was subsequently split into a training set and test set. The initial training period was between 1 October 2021 and 30 September 2022, and out of sample forecasts were made for the period 30 September 2022 to 31 December 2022. Furthermore, models were fitted for each single hour of the day and across all hours. The performance of the models was compared to a benchmark and evaluated using the root mean square error. We find that all but one model outperforms the benchmark and that modelling for each single is not better for every model. Lastly, the best model is found to be the one which combines the forecasts from the three best models based on their root mean square error. The results are consistent in some areas of electricity price forecasting, but deviates in others, therefore this thesis offers new insights to the electricity price forecasting field. | |