We have applied the algorithm to various locations for which we have data, in various parts of the world. Table 8.1 gives values we have estimated for the parameters, во and в1, and also a measure of goodness of fit of the model. We have used
Table 8.1 Parameter estimates and NRMSD for various locations
the normalised root mean square difference (NRMSD) – see Chapter 11 for the definition.
The estimates given in the table are not so dissimilar as to make us believe that we have to have a separate model for each separate location. Inspection of figures constructed using average values of the estimates of the parameters leads us to believe that there is scope for use of a so-called generic model. This model could be used to predict the diffuse radiation for any location necessary. Note that Geelong has not been included in determination of the average values. We can now use the generic model as a model for Geelong. We would argue that this is a better approach than building a separate model for Geelong anyway, since we believe that the Geelong data contains many infeasible values.
To construct the generic model, we aggregate the data from the various locations, apart from Geelong, and apply the algorithm for estimating the parameters. In this way, we obtain