Port hybrid systems#104
Conversation
Adds the bilinear (flow*head) hydro dispatch formulation and folds in adjacent refactors merged through this branch: - Hydro and storage updates to IOM helpers (#97) - POM-to-IOM type-dispatch API migration - MarketBidCost / ImportExportCost static/TS split + IEC refactor - Shiftable-load interval indexing and validation fixes - HDF system serialization (#75) - Pin GitHub revisions; bridge IOM system-query stubs to PSY public API Co-Authored-By: Luke Kiernan <86331877+luke-kiernan@users.noreply.github.com> Co-Authored-By: Rodrigo Henríquez-Auba <rodrigo.henriquezauba@nrel.gov> Co-Authored-By: Jose Daniel Lara <jdlara@berkeley.edu> Co-Authored-By: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
Remaining comments which cannot be posted as a review comment to avoid GitHub Rate Limit
JuliaFormatter
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybrid_systems.jl
Lines 819 to 820 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybrid_systems.jl
Lines 1380 to 1390 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybrid_systems.jl
Lines 1442 to 1443 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybrid_systems.jl
Lines 1483 to 1484 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybrid_systems.jl
Lines 1551 to 1553 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 132 to 134 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 183 to 184 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 191 to 192 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 211 to 214 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 216 to 217 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 223 to 224 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 252 to 253 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 296 to 297 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 303 to 304 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 321 to 324 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 326 to 327 in 4ad672f
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 332 to 333 in 4ad672f
|
Performance Results
|
There was a problem hiding this comment.
Remaining comments which cannot be posted as a review comment to avoid GitHub Rate Limit
JuliaFormatter
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 143 to 144 in e0d67dc
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 193 to 194 in e0d67dc
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 201 to 202 in e0d67dc
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 221 to 224 in e0d67dc
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 226 to 227 in e0d67dc
[JuliaFormatter] reported by reviewdog 🐶
PowerOperationsModels.jl/src/hybrid_system_models/hybridsystem_constructor.jl
Lines 233 to 234 in e0d67dc
There was a problem hiding this comment.
We need to add a couple of extra tests.
First, we need to test different attributes, that is using reservation = false, and energy target = true. Also, what happened with storage_reservation and regularization attributes?
Also, it would be good to test that the hybrid system builds when there is no thermal, or no storage, or no renewable.
Finally, it would be good to test when the hybrid does not participate in reserves.
Overall, the PR looks good. Once we have the tests, I will QA/QC the model for the different tests to ensure that the constraints are looking good.
| # Hybrid System Parameters | ||
| ################################################################################# | ||
|
|
||
| "Time-series parameter for the maximum active power available from a hybrid system's |
There was a problem hiding this comment.
@jd-lara This is a change from how things are currently defined in HSS with the time series attached to the hybrid, not the subcomponent, but I guess that works and is simpler for this case?
kdayday
left a comment
kdayday
left a comment
There was a problem hiding this comment.
Following Rodrigo's lead on the testing, but in addition, can you port over the updated equivalent of what's in the HybridDispatchWithReserves docstring on the HSS PR and have claude add hyperlinks to exported types and functions referenced in the docstring? It looks like the POM docs are way out of date, but would like to update as we go.
There was a problem hiding this comment.
Pull request overview
This PR ports/introduces PSY.HybridSystem support into PowerOperationsModels.jl by adding a new HybridDispatchWithReserves device formulation, its constructor plumbing, and a basic end-to-end test that builds/solves a hybrid co-optimization on RTS-GMLC.
Changes:
- Adds hybrid formulation types, variables/parameters/expressions/constraints, and objective plumbing for
HybridDispatchWithReserves. - Adds a two-stage
construct_device!implementation for HybridSystems (argument + model stages) and hooks it into the main module includes/exports. - Adds test utilities and a new test that builds and solves a model containing a HybridSystem with reserves.
Reviewed changes
Copilot reviewed 11 out of 11 changed files in this pull request and generated 7 comments.
Show a summary per file
| File | Description |
|---|---|
src/PowerOperationsModels.jl |
Includes hybrid model files and exports the new hybrid-related types. |
src/hybrid_system_models/hybrid_systems.jl |
Implements hybrid variable bounds, reserve aggregation, constraints, and objective-cost plumbing. |
src/hybrid_system_models/hybridsystem_constructor.jl |
Adds the two-stage construct_device! pipeline for HybridDispatchWithReserves. |
src/core/variables.jl |
Introduces HybridSystem-specific variable types. |
src/core/parameters.jl |
Introduces HybridSystem time-series parameter types and unit-conversion behavior. |
src/core/expressions.jl |
Adds HybridSystem reserve aggregation/served-reserve expression types. |
src/core/constraints.jl |
Adds HybridSystem constraint type definitions. |
src/core/formulations.jl |
Adds hybrid formulation types and user-facing documentation. |
test/test_utils/hybrid_test_utils.jl |
Adds fixtures for building a test HybridSystem in RTS-GMLC. |
test/test_device_hybrid_constructors.jl |
Adds an integration-style test that builds and solves a hybrid dispatch-with-reserves model. |
test/includes.jl |
Wires hybrid test utilities into the test suite. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
…ularization attributes; enable formatting; use jump utils; port docstrings;
Collapse paired Charge/Discharge, In/Out, Up/Down add_constraints! and
add_to_expression! methods into single bodies that dispatch on small
type-keyed trait stubs. Same JuMP shapes, same constraint meta strings,
same dispatch reachability. No public API change.
Hybrid (src/hybrid_system_models/hybrid_systems.jl):
- Charge/DischargeRegularizationConstraint
- HybridStorageStatus{Charge,Discharge}OnConstraint
- HybridStorage{Charging,Discharging}ReservePowerLimitConstraint
(folds in the _ch/_ds_reserve_up_dn_exprs helpers)
- HybridTotalReserve{Up,Down}Expression /
HybridServedReserve{Out,In}{Up,Down}Expression add_to_expression!
- ReserveAssignment/Deployment{Up,Down}{Charge,Discharge} ←
Hybrid{Charging,Discharging}ReserveVariable add_to_expression!
- HybridStatus{Out,In}OnConstraint
Storage (src/energy_storage_models/storage_models.jl):
- StorageRegularizationConstraint{Charge,Discharge}
Verified: 71/71 hybrid + storage tests pass. Net -376 lines.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replace three custom hybrid constraint types and their hand-written
add_constraints! bodies with a single call to IOM's
`add_semicontinuous_range_constraints!`, paralleling how
`AbstractThermalUnitCommitment` handles the same range-with-on-variable
pattern at thermal_generation.jl:405-419.
Mechanism:
- Define `get_min_max_limits(::PSY.HybridSystem, ::ActivePowerVariableLimitsConstraint, ::AbstractHybridFormulation)`
to read `PSY.get_active_power_limits(PSY.get_thermal_unit(d))`. IOM's
helper picks up `OnVariable` keyed by `PSY.HybridSystem` automatically.
- For the with-reserves case, introduce two expression types subtyping
`RangeConstraint{UB,LB}Expressions`: `HybridThermalActivePowerWithReserve{UB,LB}`.
Argument-stage `add_expressions!` aggregates `p_th + Σ r_up` (UB) and
`p_th − Σ r_dn` (LB) into them, after which IOM's expression-typed
dispatch emits `min·on ≤ p_th − r_dn` and `p_th + r_up ≤ max·on` directly.
Removes:
- HybridThermalOnVariableUbConstraint, HybridThermalOnVariableLbConstraint,
HybridThermalReserveLimitConstraint (constraint types + exports)
- _thermal_reserve_up_expr / _thermal_reserve_down_expr helpers
- Three add_constraints! bodies (~190 lines)
Renewable cases stay hand-written for now: IOM's parameterized helper
filters by `IS.has_time_series(d, ts_type, ts_name)`, and PSY's
HybridSystem doesn't expose its inner RenewableDispatch's time series
through that accessor. A separate change would be required.
Verified: 50/50 hybrid tests pass.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This reverts commit 59c994b.
Per PR review (acostarelli): "domain" describes the value space of each variable (a discrete set or continuous interval) more accurately than "bounds", which suggests inequality-only. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Address PR review feedback:
- "Don't use `isa`. Add a method to handle this, or restructure
existing dispatch." Eliminates every `isa(service, PSY.Reserve{...})`
and `service isa skip_kind` site in hybrid_systems.jl. The
`_excluded_reserve_kind` trait stub goes away — its information is
now encoded by union types in helper method signatures.
Five sites refactored, all sharing the same shape (per-direction
no-op methods + a fallback `::PSY.Service` work method):
- `add_to_expression!` for HybridTotalReserveExpression /
HybridServedReserveExpression → `_accumulate_reserve!`
- `add_to_expression!` for the eight Reserve*Balance{Up,Down}
{Charge,Discharge} expressions → `_balance_term!` plus a
`_deployment_factor` per-T trait that replaces the
`is_up`/`is_deployment` Booleans
- `_renewable_reserve_up/down_expr` → `_renewable_reserve_*_term!`
thunks sharing `_accumulate_renewable_reserve!`
- `_thermal_reserve_up/down_expr` → analogous restructure
- `add_constraints!` for ReserveCoverageConstraint{,EndOfPeriod}
→ `_init_coverage_container!` + `_emit_coverage_constraint!`;
the `(service isa PSY.Reserve) || continue` guard becomes the
`::PSY.Service` fallback no-op
Helper arguments use concrete types (OptimizationContainer, String,
Int, Float64, PSY.Storage, …) plus parametric `::Type{T}`/`::Type{U}`/
`d::V`/`::W` to reduce precompilation overhead.
- "Combine these if statements." Merges the two adjacent setup
ternaries (`r_ub, r_lb = if has_reserves …` and
`con_lb = if has_reserves …`) in
`add_constraints!(::Type{HybridStatus{Out,In}OnConstraint}, …)` into
a single 3-tuple ternary.
Test.detect_ambiguities returns 0; full suite passes (50/50).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…milies
Hybrid reserve variables, expressions, and constraints had ~16 paired Charge/Discharge,
Up/Down, Total/Served (a.k.a. Assignment/Deployment), and UB/LB singleton structs plus
~14 paired trait helpers that all differed only by which sibling they referenced.
Introduce marker singletons for ReserveSide, ReserveDirection, ReserveScale (UnscaledReserve
/ DeployedReserve), reuse IOM's BoundDirection (UpperBound/LowerBound), and reparametrize
the family roots:
- ReserveAggregationExpression{D,S,Sd} umbrella with two concrete struct families
(HybridPCCReserveExpression, StorageReserveBalanceExpression) covering all 16
historical reserve-expression singletons.
- HybridPCCReserveVariable{Sd}, HybridStorageSubcomponentReserveVariable{Sd},
HybridStorageSubcomponentPower{Sd}, RegularizationVariable{Sd}.
- HybridStatusOnConstraint{Sd}, HybridStorageStatusOnConstraint{Sd},
HybridStorageReservePowerLimitConstraint{Sd}, RegularizationConstraint{Sd},
HybridThermalOnVariableConstraint{B}.
All 34 historical concrete names are retained as const aliases so external imports,
`get_expression`/`get_variable`/`get_constraint` lookups, and module exports are
byte-compatible. Inside hybrid_systems.jl this lets:
- _accumulate_reserve! + _balance_term! collapse into one _add_reserve_term! family
(PCC boundary and storage subcomponent share the no-op skip and scale dispatch);
- thermal/renewable subcomponent accumulators (10 helpers) collapse into one
_subcomponent_reserve_term! / _subcomponent_reserve_expr family parametric on
the variable type;
- the UB/LB thermal-on-variable add_constraints! methods merge;
- ~14 paired trait helpers (storage / PCC / regularization) become parametric
single-method definitions;
- 5 file-local Union consts (_BalanceUpExpr, _BalanceDownExpr, _BalanceDeploymentExpr,
_HybridReserveUpExpr, _HybridReserveDownExpr) and _StorageCharge/DischargeSide
Union consts get deleted.
Additional cleanups:
- get_variable_multiplier hybrid signatures take ::Type{<:Formulation} (matches the
rest of POM); all W() instance call-sites become type-keyed.
- Three `if W <: ...` body-level subtype checks split into separate parametric
dispatched methods (HybridStorageBalanceConstraint, RegularizationConstraint,
HybridStatusOnConstraint).
- _init_coverage_container! uses lazy_container_addition! (idempotent).
- add_proportional_cost!(OnVariable, hybrids) hoists the variant/invariant
function-handle selection out of the per-t loop.
Net: -142 lines across 5 files. Full Pkg.test passes (13125 / 0 fail / 0 error / 1
pre-existing broken). Zero method ambiguities.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Port QuadraticLossConverterMILP and HVDCTwoTerminalVSCLP from a hardcoded SOS2 depth (DEFAULT_INTERPOLATION_LENGTH) to the same tolerance/attribute API used by HydroTurbineMILPBilinearDispatch: "bilinear_approximation", "bilinear_quadratic_method", "bilinear_tolerance". All five schemes are supported. The squares-based schemes (bin2, hybs, none) reuse the standalone loss i_sq via IOM's precomputed (xsq, ysq) overload; the discretization-based schemes (nmdt, dnmdt) never build i², so they take the raw form with nothing to duplicate. The precomputed-vs-raw branch is centralized in _add_converter_bilinear!, and config construction dispatches on the formulation type so the *NLP types stay exact. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The single absolute bilinear_tolerance was an absolute gap on the v·I / flow·head product, whose magnitude differs by formulation, so the same 1e-2 was far stricter for converters (depth 15, intractable) than for hydro. Replace it across all bilinear formulations (HydroTurbineMILPBilinearDispatch, QuadraticLossConverterMILP, HVDCTwoTerminalVSCLP) with two keys: bilinear_relative_tolerance (default 0.05, a fraction of the product magnitude and the default sizing knob) and bilinear_absolute_tolerance (optional). A relative tolerance is scaled to absolute by the term magnitude via the new _resolve_tolerance / _max_abs helpers (max|x|·max|y| for the bilinear, max|i|² for the standalone I² loss term); when both are set the finer binds. The gap→depth inversion stays in IOM; POM does the relative→absolute scaling since it needs the bounds. Default depth drops from 15 to 5 on the converter test systems. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Collapse each NLP/MILP formulation pair into one formulation whose "bilinear_approximation" attribute defaults to the exact "none" case (IOM's NoApproximation configs), opting into MILP via a linearizing scheme: - HydroTurbineMILPBilinearDispatch -> HydroTurbineBilinearDispatch - QuadraticLossConverterMILP/NLP -> QuadraticLossConverter - HVDCTwoTerminalVSCLP/NLP -> HVDCTwoTerminalVSC The VSC PQ-capability (exact disk vs octagon) and pq-square registration are re-keyed from formulation-type dispatch to dispatch on the IOM bilinear config type, keeping the exact/approximate split branch-free. Old *MILP/*NLP/*LP names are removed (no aliases). Tests updated to select the MILP path via an explicit attribute. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…rename VSC helpers - Add shared BILINEAR_APPROX_DEFAULT_ATTRIBUTES constant (single source of the MILP approximation defaults + their documentation); merge it into the hydro, QuadraticLossConverter, and VSC get_default_attributes instead of duplicating. - Shorten the three formulation docstrings to reference the constant. - _resolve_tolerance now requires exactly one of absolute/relative (error on both or neither) instead of silently taking the min. - Rename the cryptic VSC pq/_capability helpers to apparent-power-limit names (matching HVDCVSCApparentPowerLimitConstraint); update call sites. - Drop three stale comments. - Trim HVDC tests: remove the pure-construction config-bridge testset (replaced by a focused tolerance check), coarsen the MILP solve models, and cover only representative bilinear schemes (bin2 + nmdt). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The MT-HVDC "QuadraticLossConverter agreement" test compared only the objective, which is dominated by generation cost while the converters carry ~zero current at the optimum (the default CopperPlatePowerModel collapses the two AC islands the DC link bridges, so flow is never needed). The assertion passed vacuously and the accompanying comment rationalized it incorrectly. - Replace it with a conservativeness bound (milp_obj <= nlp_obj * 1.06): the bin2 McCormick relaxation lower-bounds the NLP, allowing for the 5% MIP gap. Use horizon=3h + mip_rel_gap=0.05 so the SOS2 model solves in ~1s instead of timing out. Add a TODO documenting why forced flow is currently unbuildable (DCPPowerModel + VoltageDispatchHVDCNetworkModel fails: QuadraticLossConverter wires into ActivePowerBalance__DCBus, which only the copperplate path creates). - Strengthen the VSC test (genuine forced flow: the VSC replaces a line) to assert the solutions agree, not just objectives: both models push the VSC past 1.5 pu and aggregate throughput agrees within rtol 0.1. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ids, SOC target
- Fix undefined `Up`/`Down` type params in storage reserve expressions
(StorageReserveBalanceExpression{Up/Down,...} -> {PSY.ReserveUp/ReserveDown,...}),
which previously errored when constructing storage ancillary services.
- C4: create TotalReserveOffering containers for every hybrid that participates in a
reserve service, not just hybrids with storage. get_expression_type_for_reserve
routes all hybrids' ActivePowerReserveVariable into TotalReserveOffering, so
storage-less hybrids with reserves no longer hit a missing-container error. The
subcomponent feed stays gated on storage. Adds a regression test.
- C7: give the hybrid end-of-period energy target its own HybridEnergyTargetConstraint
(a one-sided floor e_T >= E_T, no slacks) instead of reusing the storage
StateofChargeTargetConstraint (an equality with surplus/shortfall slacks), and fix
the hybrid formulation docstring to match.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ity) The HybridDispatchWithReserves port of the end-of-period storage energy target was mis-ported from HybridSystemsSimulations.jl: it dropped the surplus/shortage slack variables and implemented a hard one-sided floor (e_T >= E_T) instead of HSS's soft equality with penalized slacks. The energy_target path was never exercised by any test, so this went unnoticed (and the slack-typed add_variables!/add_constraints! signatures only accepted FlattenIteratorWrapper, never the Vector the constructor passes, so the constraint method never even matched). Mirror POM's storage StateofChargeTargetConstraint, adapted for the hybrid: - Add HybridEnergySurplusVariable / HybridEnergyShortageVariable (non-negative, final-time-step only) and export them. - Make HybridEnergyTargetConstraint a soft equality e_T - e+ + e- = E_T. - Penalize both slacks in the objective from the storage subcomponent's StorageCost (energy_surplus_cost / energy_shortage_cost), gated on energy_target. - Add the slacks in the constructor ArgumentConstructStage. - Broaden the slack add_variables! and the target add_constraints! to accept Vector as well as FlattenIteratorWrapper. Keep the existing target RHS scaling (storage_target is a ratio of capacity in PSY; the hybrid EnergyVariable is absolute energy), which is the one intentional divergence from HSS's raw get_storage_target. Add tests modeled on the storage energy-target tests: assert the slacks exist and the constraint is an equality (would have caught the regression), that the slacks are absent when energy_target=false, and an on-vs-off objective check confirming the penalty reaches the objective. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
| con = get_constraint(container, T, V, "$(s_type)_$(s_name)_discharge") | ||
| jm = get_jump_model(container) | ||
| if time_offset(T) == -1 | ||
| con[ci_name, 1] = JuMP.@constraint( | ||
| jm, | ||
| sustained_param_discharge * reserve_var[ci_name, 1] <= get_value(ic) | ||
| ) | ||
| for t in time_steps[2:end] | ||
| con[ci_name, t] = JuMP.@constraint( | ||
| jm, | ||
| sustained_param_discharge * reserve_var[ci_name, t] <= | ||
| energy_var[ci_name, t - 1] | ||
| ) | ||
| end | ||
| else # EndOfPeriod | ||
| for t in time_steps | ||
| con[ci_name, t] = JuMP.@constraint( | ||
| jm, | ||
| sustained_param_discharge * reserve_var[ci_name, t] <= | ||
| energy_var[ci_name, t] | ||
| ) | ||
| end |
There was a problem hiding this comment.
@rodrigomha @jd-lara What do you think of this? I accepted the change, but this isn't how HSS does it.
| [sources] | ||
| InfrastructureSystems = {rev = "IS4", url = "https://github.com/Sienna-Platform/InfrastructureSystems.jl"} | ||
| PowerSystems = {rev = "psy6", url = "https://github.com/Sienna-Platform/PowerSystems.jl"} | ||
| # TODO: Move to main once this branch is merged | ||
| PowerSystems = {rev = "ac/hybridsystem-strip-units", url = "https://github.com/Sienna-Platform/PowerSystems.jl"} | ||
| InfrastructureOptimizationModels = {rev = "main", url = "https://github.com/Sienna-Platform/InfrastructureOptimizationModels.jl"} | ||
| PowerSystemCaseBuilder = {rev = "psy6", url = "https://github.com/Sienna-Platform/PowerSystemCaseBuilder.jl"} | ||
| PowerNetworkMatrices = {rev = "psy6", url = "https://github.com/Sienna-Platform/PowerNetworkMatrices.jl"} |
| PowerOperationsModels = "bed98974-b02a-5e2f-9ee0-a103f5c450dd" | ||
| PowerSystemCaseBuilder = "f00506e0-b84f-492a-93c2-c0a9afc4364e" | ||
| PowerSystems = "bcd98974-b02a-5e2f-9ee0-a103f5c450dd" | ||
| PrettyTables = "08abe8d2-0d0c-5749-adfa-8a2ac140af0d" |
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Thanks for opening a PR to PowerOperationsModels.jl, please take note of the following when making a PR:
Check the contributor guidelines