Well primaryvariable scaling

[hnil]

• make scaling of primary variables of wells
• scale control equation if it is bhp control
  Idea is to be able to make scaling of the matrices more normal. It is a follow up on the scaling of the reservoir primary variables.

[GitPaean]

jenkins build this please

[GitPaean]

benchmark please

[ytelses]

Benchmark result overview:

Test	Configuration	Relative
opm-git	OPM Benchmark: flow_mpi_extra - Threads: 1	1.004
opm-git	OPM Benchmark: flow_mpi_extra - Threads: 8	0.998
opm-git	OPM Benchmark: flow_mpi_norne - Threads: 1	0.999
opm-git	OPM Benchmark: flow_mpi_norne - Threads: 8	0.991
opm-git	OPM Benchmark: flow_mpi_norne_4c_msw - Threads: 1	0.997
opm-git	OPM Benchmark: flow_mpi_norne_4c_msw - Threads: 8	0.996

• Speed-up = Total time master / Total time pull request. Above 1.0 is an improvement. *

View result details @ https://www.ytelses.com/opm/?page=result&id=1830

[GitPaean]

Do you have some ideas about what kind of number should be used to benefit a typical blackoil simulation running?

[GitPaean]

Should bhp_scaling and bhp_control_scaling be same or same order or something very different?

[GitPaean]

From here, if the bhp_scale_ and bhp_control_scale are different values, how will it affects the tolerance checking?

[GitPaean]

hei, one small thing is that, for the bhp_scaling, they are used to scale to reduce the values for the related primary variables through division, while for bhp_control_scaling, it is used to reduce the orders of the control_eq through multiplication. As a result, bhp_scaling will be a big value maybe in the order or bar level, while bhp_control_scaling will be a very small value.

Do you think it is easier to use if we can change this to make both bhp_scaling and bhp_control_scaling to be big values?

[hnil]

Closing: five of the seventeen touched files no longer exist after the Ebos rename, and the well primary-variable code has moved on considerably since 2022. Well primary-variable scaling is better approached against the current StandardWellPrimaryVariables. Branch retained on my fork.

[hnil]

Reopening as a draft to keep it alive as a marker rather than lose the topic. Well primary-variable scaling is still worth reviewing at some point.

To be clear about its state: this cannot be rebased as-is -- five of the seventeen touched files no longer exist after the Ebos rename. Treat it as a record of the idea, to be rewritten against the current StandardWellPrimaryVariables when someone picks it up.

[hnil]

Superseded by #7286, which is the same idea done properly: the scaling is applied to the AD derivative only, so stored values stay physical and B/C are untouched, and it comes with measured conditioning numbers for D. Branch kept.