Improve precision of BICUBIC interpolation close to the critical point

[casella]

I am involved in several studies of sCO2 systems, for which the interpolated CO2 medium of CoolProp is extremely handy. However, sCO2 systems exploit the strong real-gas effects close to the critical point, so it is important that the fluid properties are computed with a decent approximation in the region just above the critical point in the ph-plane.

The attached model TestInterpolatedCO2.mo.txt computes the fluid properties on an isobaric line at 90 bar, which is relatively close to the critical pressure (but not too close either), using the setState_pT function.

These two plots show the comparison between the full CoolProp model and the BICUBIC interpolation, when giving pT as input and retrieving density and specific enthalpy


The maximum error on density is about 8%, which is not dramatic, but still significant.

Is there a simple way to increase the number of data points of the interpolation table, by just adding stuff to the medium configuration string? I understand doubling the number of nodes on both axes would increase the table size by a factor four, but give the cubic approximation, the error will shrink by 2^4, i.e. become less than 0.5%.

It could also be useful if one could specify the upper and lower limits of pressure, cutting out useless data points that will never be used for the application at hand (e.g. 74 and 300 bars for sCO2 cycles).