Let reservoir coupling slave UDQs see the production limits in force

[hakonhagland]

Replaces #7465, as agreed there: instead of stopping a slave whose UDQs use GOPRT and the other production target vectors for a slave group, report the right number. This is the production counterpart of #7413, which did this for GGIRT/GWIRT.

Depends on OPM/opm-common#5421, which adds data::ReservoirCouplingGroupRates::production_targets and lets the GOPRT, GWPRT, GGPRT and GLPRT evaluators use it.

Report the production limits in force for slave groups (commit 1)

• The master sends a slave group a production limit for every rate type: the target of its active control mode, and its guide-rate share of the limits further up the master's group tree (GroupConstraintCalculator::groupProductionConstraints()). A slave UDQ that refers to GOPRT, GWPRT, GGPRT or GLPRT for such a group nevertheless read the slave deck's own GCONPROD value, usually zero.
• New BlackoilWellModelRescoup::storeSlaveGroupProductionTargets(), next to storeSlaveGroupInjectionTargets(). It works out the limit in force per slave group and rate type: the master's, the deck's own, or the smaller of the two, as the group's GRUPSLAV flag says. It then:
  • keeps it on the slave (ReservoirCouplingSlave::effectiveProductionTargets()), so that EclWriter passes it to the summary output through production_targets;
  • primes the summary state with it before the group and field UDQs are evaluated. When nothing is in force, it primes the schedule's own limit.
• It runs at the same two points as the injection targets: after the handshake at the start of a sync step, and after the mid-step receive. refreshSlaveGroupInjectionTargets() is renamed to refreshSlaveGroupTargets(), since it now refreshes both.
• The rule is the one GroupStateHelper::getEffectiveProductionLimit_() applies in the control path, with one difference. There the deck's own limit is only consulted for rate types the group has its own GCONPROD limit for. A limit is reported here for every rate type, so under BOTH the deck's own limit only takes part when the group has one, as effectiveSlaveGroupInjectionTarget_() does for injection.
• RESV is left out: GVPRT does not report a group limit, but adds up the reservoir volume rate targets of the group's wells. GVPRT, GVIRT, the master's GVPR/GVIR and gas lift are to be handled in later PRs, as discussed in Stop a reservoir coupling slave whose UDQs use a target vector it cannot report #7465.

Testing

Two runs of a three-model coupled case: a master and two slaves, METRIC units, master GCONPROD with ORAT targets and a gas limit of 6.9E6 on the master groups.

• Unchanged case: summary output in the master and both slaves is identical to a run built from an earlier master, at zero tolerance (it reads none of the four vectors).
• The same case, with the slave's SUMMARY asking for GOPRT, GWPRT, GGPRT and GLPRT for all groups, and a UDQ FUOPTC = GOPRT of one slave group:
  • each producing slave group's GOPRT equals the master's GOPR for its master group at every report step (the field's 9000 split by guide rate). Without this PR it reads the slave deck's own value, which is 0 here, since the slave deck has no GCONPROD;
  • the two producing slave groups' GGPRT add up to the master's 6.9E6 gas limit;
  • GWPRT and GLPRT stay 0: the master sets no water or liquid limit;
  • the UDQ equals that group's GOPRT at all 59 timesteps, including the step where the target changes, so it does not lag a step behind.

[hakonhagland]

Putting this in draft mode until OPM/opm-common#5421 has been merged

[Copilot]

Copilot review overview

🟡 Changes recommended

An empty production handshake can leave stale limits that are subsequently reported as current targets.

Review effort: Balanced
Findings: 1

Open (1)

• Clear stale production limits on empty handshake · New

What changed in this PR

Adds effective slave-group production limits so reservoir-coupling UDQs and summary vectors report master-imposed values.

Changes:

• Computes effective GOPRT/GWPRT/GGPRT/GLPRT limits.
• Refreshes target-dependent UDQs during synchronization.
• Passes production limits to the opm-common summary evaluators.

File	Description
BlackoilWellModelRescoup.hpp	Declares production-target handling.
BlackoilWellModelRescoup_impl.hpp	Computes, stores, and primes effective limits.
BlackoilWellModel_impl.hpp	Invokes the expanded target refresh.
ReservoirCouplingSlave.hpp	Stores effective production targets.
EclWriter.hpp	Exposes targets to summary evaluation.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[hakonhagland]

jenkins build this serial please

[totto82]

Thanks. The opm-common part is already merged. The code looks correct.