Model inflexible devices as assets; drop the unreleased flex-context group field

[Flix6x]

Description

Changes how inflexible (unschedulable) devices are modelled, and removes an unreleased sub-feature of #2358.

An inflexible device can now be declared in a flex-model entry (as its own asset) by giving it a single inflexible-consumption or inflexible-production sensor reference (the field name sets the sign; source filters are supported). Such an entry carries no schedulable-device fields; it simply declares a fixed device whose power is accounted for. It joins a group through the ordinary flex-model group field, exactly like a flexible member — so:

• inflexible devices are no longer an exception: group membership is defined the same way for every element (the element points at its group);
• the field is a single reference, not a list;
• inflexible devices are gathered upward with the rest of the flex-model / asset tree, so we never need to gather flex-context downward into a subtree (keeping the invariant "a flex-context only ever inherits downward").

The flat inflexible-consumption / inflexible-production lists in the flex-context remain, for plain site base load (no group).

Removed (unreleased, from #2358)

• The group field on flex-context inflexible entries, the SensorReference.group carrier, and the device-inventory plumbing behind it (_collect_inflexible_referenced_group_keys, FlexDevice.group_reference, the inflexible branch of group_key). group_to_devices still folds inflexible members in, and by_index stays — both are reused by the new model.

Notable

• InflexibleDeviceSchema moves to schemas/sensors.py so both the flex-context and the flex-model can use it without an import cycle.
• Validation on a flex-model inflexible entry: rejects declaring both signs, a sensor whose explicit consumption_is_positive contradicts the field, and co-existing schedulable-device fields.
• Canonical solver order: flexible devices → flex-model inflexible devices → flat flex-context inflexible devices.

How to test

• pytest flexmeasures/data/models/planning/tests/test_device_inventory.py (classification + ordering + group membership)
• pytest flexmeasures/data/models/planning/tests/test_group_constraints.py -k inflexible (an inflexible load in a group caps the battery end-to-end)
• pytest flexmeasures/data/schemas/tests/test_scheduling.py -k "inflexible or storage_flex_model" (schema validation)

Note: test_battery_storage_with_time_series_in_flex_model fails on this branch and on clean main in this environment (a price-fixture/shared-DB artifact, UnknownPricesException); unrelated to this change.

Related

• Supersedes the flex-context group-on-inflexible sub-feature of Replace flex-context field inflexible-device-sensors with inflexible-consumption and inflexible-production #2358 (unreleased). Part of the Support intermediate power constraints on groups of devices (flex-model group field) #2276 group work (point 5 of that discussion).

🤖 Generated with Claude Code

[read-the-docs-community]

Documentation build overview

📚 flexmeasures | 🛠️ Build #33857916 | 📁 Comparing a95773b against latest (26d71eb)

  🔍 Preview build  

10 files changed · ± 10 modified

± Modified

• changelog.html
• genindex.html
• _autodoc/INFLEXIBLE_CONSUMPTION.html
• _autodoc/INFLEXIBLE_PRODUCTION.html
• _autosummary/flexmeasures.data.models.planning.devices.html
• _autosummary/flexmeasures.data.schemas.scheduling.html
• _autosummary/flexmeasures.data.schemas.scheduling.storage.html
• _autosummary/flexmeasures.data.schemas.sensors.html
• api/v3_0.html
• features/scheduling.html

[Copilot]

🟡 Not ready to approve

The OpenAPI schema/examples for the new flex-model inflexible fields currently document an array example for fields that accept a single object, which can mislead API/UI users.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

This review doesn't count toward merge requirements. Sign up for the private preview to control whether Copilot approvals count.

Pull request overview

This PR updates FlexMeasures’ scheduling model to treat inflexible (unschedulable) devices as assets declared in the flex-model (via a single inflexible-consumption or inflexible-production sensor reference), and removes the previously introduced but unreleased flex-context inflexible group sub-feature. This unifies group membership semantics (everything joins groups via the normal flex-model group field) and simplifies device-inventory classification and solver ordering.

Changes:

• Add flex-model support for inflexible devices (inflexible-consumption / inflexible-production) with schema-level validation to prevent ambiguous/invalid entries.
• Update device inventory classification/enumeration so flex-model inflexible entries are indexed after flexible devices and before flat flex-context inflexibles, and so only flex-model entries can participate in groups.
• Update docs/tests/OpenAPI to reflect the new modeling approach and the removal of the flex-context inflexible group carrier.

File summaries

File	Description
flexmeasures/ui/static/openapi-specs.json	Updates generated OpenAPI schema for removed flex-context group support and added flex-model inflexible fields (but currently contains misleading examples for these new fields).
flexmeasures/data/schemas/tests/test_scheduling.py	Replaces flex-context-group validation test with flex-model inflexible-field validation tests.
flexmeasures/data/schemas/sensors.py	Moves InflexibleDeviceSchema into a shared location and removes SensorReference.group.
flexmeasures/data/schemas/scheduling/storage.py	Adds flex-model fields for inflexible devices plus schema validation to enforce correct/unambiguous declarations.
flexmeasures/data/schemas/scheduling/metadata.py	Updates descriptions of inflexible fields to reflect “list in flex-context vs single ref in flex-model entry”.
flexmeasures/data/schemas/scheduling/init.py	Removes old scheduling-local InflexibleDeviceSchema (with group) and updates UI schema metadata for inflexible fields.
flexmeasures/data/models/planning/tests/test_group_constraints.py	Updates group-constraint test to model an inflexible load as its own flex-model entry/asset and join group via normal group.
flexmeasures/data/models/planning/tests/test_device_inventory.py	Updates inventory tests to cover new inflexible classification, ordering, and group membership semantics.
flexmeasures/data/models/planning/devices.py	Implements flex-model inflexible-entry classification and updates group membership resolution (flex-model only).
documentation/features/scheduling.rst	Documents flex-model inflexible fields and explains new approach for inflexible devices participating in groups.
documentation/changelog.rst	Adds changelog entry for modeling inflexible devices as assets and removes mention of unreleased flex-context group-on-inflexible feature.

Review details

Suppressed comments (3)

flexmeasures/data/schemas/scheduling/storage.py:381

• The inflexible-production field on a flex-model entry is a single nested reference, but it reuses metadata.INFLEXIBLE_PRODUCTION whose example is a list; this makes the generated OpenAPI/Docs misleading (array vs. object).

    inflexible_production = fields.Nested(
        InflexibleDeviceSchema,
        data_key="inflexible-production",
        required=False,
        metadata=metadata.INFLEXIBLE_PRODUCTION.to_dict(),

flexmeasures/data/schemas/scheduling/storage.py:664

• Same issue as above for DB-stored flex-model entries: inflexible-production is a single nested object, but the reused metadata example is a list, which misdocuments the expected input shape.

    inflexible_production = fields.Nested(
        InflexibleDeviceSchema,
        data_key="inflexible-production",
        required=False,
        metadata=metadata.INFLEXIBLE_PRODUCTION.to_dict(),

flexmeasures/ui/static/openapi-specs.json:6870

• In the generated OpenAPI schema for StorageFlexModelSchemaOpenAPI, inflexible-production is a single object ($ref to InflexibleDevice), but the example is an array. This is inconsistent with the actual schema and may mislead API/UI users.

            "example": [
              {
                "sensor": 3
              },
              {

• Files reviewed: 11/11 changed files
• Comments generated: 3
• Review effort level: Lite

We're testing this review assessment. Please use 👍 or 👎 to tell us if it's correct.

[Flix6x]

Fixed in c4de6dd: the flex-model inflexible-consumption/inflexible-production fields now override the shared metadata's list example with a single object ({"sensor": 3}), so the generated OpenAPI documents an object ($ref InflexibleDevice), not an array. The flex-context list fields keep their array example.

[Flix6x]

Fixed in c4de6dd: the flex-model inflexible-consumption/inflexible-production fields now override the shared metadata's list example with a single object ({"sensor": 3}), so the generated OpenAPI documents an object ($ref InflexibleDevice), not an array. The flex-context list fields keep their array example.

[Copilot]

🟡 Not ready to approve

The new validation intended to forbid schedulable-device fields on inflexible flex-model entries currently misses at least soc-unit, allowing contradictory inputs to pass.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

This review doesn't count toward merge requirements. Sign up for the private preview to control whether Copilot approvals count.

Review details

Suppressed comments (1)

flexmeasures/data/schemas/scheduling/storage.py:57

• validate_inflexible_flex_model_entry is meant to reject any schedulable-device fields on an inflexible flex-model entry, but _SCHEDULABLE_DEVICE_DATA_KEYS currently omits soc-unit. This allows inputs like { "inflexible-consumption": {…}, "soc-unit": "kWh" } to pass validation even though the entry is still treated as inflexible, which contradicts the stated validation rules and can lead to silently ignored fields.

    "soc-maxima",
    "soc-targets",
    "soc-gain",
    "soc-usage",

• Files reviewed: 11/11 changed files
• Comments generated: 0 new
• Review effort level: Lite

We're testing this review assessment. Please use 👍 or 👎 to tell us if it's correct.

[Flix6x]

Addressed Copilot's second-round finding in fd78a771a: the schedulable-field check on an inflexible flex-model entry used a hand-maintained blacklist that missed soc-unit. Replaced it with a whitelist of the keys allowed alongside an inflexible declaration (identity/group/commodity + the defaulted activation preferences), so every other declared field — including soc-unit and any future device field — is rejected. Added a soc-unit test guard.

[Copilot]

🟡 Not ready to approve

A small docstring-formatting fix is needed and the new “source-filtered inflexible device in flex-model” path lacks a dedicated regression test.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

This review doesn't count toward merge requirements. Sign up for the private preview to control whether Copilot approvals count.

Review details

Suppressed comments (2)

flexmeasures/data/schemas/sensors.py:1060

• The new InflexibleDeviceSchema docstring introduces line breaks mid-sentence (e.g. breaking after “in a”), which conflicts with the repository docstring/comment convention to break lines only after punctuation for readability and easier searching.

    """One inflexible device: a sensor reference with optional source filters.

    Used both in the flex-context (as a list, for site-level inflexible load) and in a
    flex-model entry (as a single reference, when an inflexible device is modelled as
    its own asset). Deserializes to a plain :class:`Sensor` when no source filters are
    given (a backward-compatible shape downstream) and to a :class:`SensorReference`
    otherwise.
    """

flexmeasures/data/models/planning/devices.py:343

• Support for source-filtered inflexible devices in the flex-model is implemented by accepting SensorReference-like objects (entry_ref.sensor) here, but there is no test covering that path. Adding a regression test would ensure source filters are preserved on FlexDevice.sensor_reference for flex-model-declared inflexible entries (not just the flat flex-context list).

        # Tolerate SensorReference-like objects (with a `.sensor`) and plain Sensors.
        reference_sensor = getattr(entry_ref, "sensor", None)
        if reference_sensor is not None:
            sensor, sensor_reference = reference_sensor, entry_ref
        else:
            sensor, sensor_reference = entry_ref, None
        return FlexDevice(

• Files reviewed: 11/11 changed files
• Comments generated: 0 new
• Review effort level: Lite

We're testing this review assessment. Please use 👍 or 👎 to tell us if it's correct.

[Flix6x]

Addressed round 3 in the latest commit: reflowed the new docstrings to break only after punctuation (matching the repo convention / surrounding code), and added a device-inventory regression test that a flex-model inflexible entry given as a source-filtered SensorReference preserves the reference on FlexDevice.sensor_reference (so source filters reach the solver's power lookup).

[Copilot]

🟡 Not ready to approve

A single-sensor flex-model dict containing inflexible-consumption/inflexible-production can be accepted by schema deserialization but will not be classified as an inflexible entry in DeviceInventory, risking silent misconfiguration.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

This review doesn't count toward merge requirements. Sign up for the private preview to control whether Copilot approvals count.

Review details

Suppressed comments (1)

flexmeasures/data/models/planning/devices.py:448

• In single-sensor mode (flex_model is a dict), flex-model-declared inflexible devices are not classified (_register_flex_model_inflexible is a no-op), but the schema layer can still deserialize a dict that only contains inflexible-consumption/inflexible-production. That would currently be treated as a schedulable DEVICE entry (using the target sensor) rather than being rejected, which can silently accept an invalid configuration.

        # Collect the group keys referenced by flex-model entries' "group" fields; the
        # entries whose own sensor/asset matches a referenced key are classified as group
        # entries below. Flex-model-sourced inflexible devices carry their "group" the
        # same way, so this one scan covers them too.
        inventory.referenced_group_keys = _collect_referenced_group_keys(

• Files reviewed: 11/11 changed files
• Comments generated: 0 new
• Review effort level: Lite

We're testing this review assessment. Please use 👍 or 👎 to tell us if it's correct.

[Flix6x]

Addressed round 4: a single-sensor flex-model declaring inflexible-consumption/inflexible-production now raises "only supported in multi-device flex-models" (mirroring the group field) instead of silently being scheduled as a normal device. Added a regression test.

[Copilot]

🟡 Human review recommended

It changes core scheduling/device-inventory classification and grouping semantics, so a final human review should confirm no subtle regressions beyond the updated tests and docs.

Review details

• Files reviewed: 11/11 changed files
• Comments generated: 0 new
• Review effort level: Lite

We're testing this review assessment. Please use 👍 or 👎 to tell us if it's correct.

[Flix6x]

Suggested change

* An inflexible (unschedulable) device can be modelled as its own asset by giving its flex-model entry a single ``inflexible-consumption`` or ``inflexible-production`` sensor reference; such a device joins a ``group`` like any other member, so its fixed (measured) load counts towards the group's intermediate power constraint [see `PR #2374 <https://www.github.com/FlexMeasures/flexmeasures/pull/2374>`_]

* An inflexible (unschedulable) device can be modelled as its own asset by giving its flex-model entry a single ``inflexible-consumption`` or ``inflexible-production`` sensor reference; such a device can be assigned to a ``group`` like any other member, so its fixed (measured) load counts towards the group's intermediate power constraint [see `PR #2374 <https://www.github.com/FlexMeasures/flexmeasures/pull/2374>`_]