Fig. 3 shows a selection of parameter values for this TH-302 exploratory analysis. While this figure shows results for each parameter keeping all other parameters fixed to their optimum value, in most cases all possible combinations of all parameters were actually executed (not shown). All but one parameter (the analog trend) were found to be site independent. While one value may sometimes seem more appropriate for a given case, each optimal value was also selected for its pertinence to the overall improvement of the results (both deterministic and probabilistic, and across all cases tested).
Fig. 3. Sensitivity analysis to the number of analogs, analog trend, analog predictors, and training period. Results are shown in terms of Spearman rank correlation coefficient, centered root-mean-square error (CRMSE) and bias. For each panel, all other analog parameters are fixed to their optimum values. The vertical dashed line indicates the chosen optimum value(s) of the associated parameter. The following analog predictors were tested: wind speed (ws), wind direction (wd), surface pressure (p), planetary boundary layer height (pb), specific humidity (q), and/or temperature (t). The training period is given as a number of days or a year.Figure optionsDownload full-size imageDownload as PowerPoint slide