Open Access
Table 1
Summary of the methodology used for the forecasting and clustering analysis.
| Step | ANN | SVR and LS-SVR | Clustering and forecast |
|---|---|---|---|
| 1 | Import and clean raw data from missing points and outliers | ||
| 2 | Choose a resolution and horizon | Resolution: 60 min, horizon:24 h | |
| 3 | Organize input matrix X and output vector Y | ||
| 4 | Normalize X and Y | Organize Y as 24 hourly daily and re-arrange X accordingly | |
| 5 | Randomly partition data into 10 folds | Use daily variance and peak index values to be used for clustering | |
| 6 | Allocate 9 folds for training and validation of the models | Run K-means with optimum number of clusters | |
| 7 | Tune the ANN ensemble parameters using the training set | Tune the SVR and LS-SVR parameters using the training set | Assign daily profiles to the relevant cluster |
| 8 | Choose optimum parameters according to the smallest 10 fold cross-validation error | Repeat steps 4–8 of ANN,SVR and LS-SVR methods on the clusters | |
| 9 | Test the models on the independent test set and store the results in the “Predictions” vector | Store results separately for each cluster and also for the entire clustered data in Predictions_wCluster | |
| 10 | Repeat the first 9 steps for each of the 10 folds. | Calculate error metrics and compare the results obtained by Predictions_wCluster and Predictions | |
| 11 | Calculate error metrics by using the “Predictions” vector and the corresponding real load values | ||
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.