New aggregation control to design calibration

Normally a calibration using Multirun options are made using New candidates  of parameter values for each run.

However, in case of trying to find common parameter values for various runs as represented by a deterministic change of  som paramters and stochastic change of other parameters a new procedure has been developed.

The use of this is recommended together with the assignment of Multirun parameters into different dimenension.

Parameters in the first dimension are expected to change by a fixed pattern and we can recommend the Table method to assign exact values for each of those parameters.

Parameters in the second dimension are expected for parameters that are to be calibrated by GLUE or Bayesian methods.

The number of repetitions within the first dimension will then be possible to force into a single performance for all the runs within the first dimension for the single change of parameters in the second dimension. To force the calibration to provide a common value for the number of repetitions within the first dimension a New Technical Parameter should be assigned to a value that corresponds to the number of repetions within that dimension.

Then new parameter (Technical Model” is named: Aggregate Calib Number. The default value is 1- meaning that no aggregation of performance will be made.

In addition a new parameter is introduced to facilitate the use of measured data from more that one site.

The new parameter is called: Index in val file.

The parameter is enabled by a switch named: IndexForValidation. By default this index is selected from the when validation variables are assigned to match simulated variables. However if the option : From serie is selected the index will be controlled by values of the index in val file parameter.





Leave a Reply

Your email address will not be published. Required fields are marked *