Reset components to default power when all proposals are withdrawn #1212
Conversation
The power manager in some cases keeps resending the bounds even when there are no changes. And these tests were relying on that behaviour. This is changed now, so that the power manager's behaviour can change as well. Signed-off-by: Sahas Subramanian <[email protected]>
Signed-off-by: Sahas Subramanian <[email protected]>
With this, the power manager algorithms basically start ignoring the `must_send` flag and always return a target power value if available. Signed-off-by: Sahas Subramanian <[email protected]>
Signed-off-by: Sahas Subramanian <[email protected]>
Signed-off-by: Sahas Subramanian <[email protected]>
Signed-off-by: Sahas Subramanian <[email protected]>
github-actions
bot
added
part:tests
There was a problem hiding this comment.
Pull Request Overview
This PR changes the behavior of power proposals so that when all proposals for a set of components are withdrawn (i.e. when propose_power is set to None), the components are reset to their respective default powers as defined by the component category. Key changes include updating the tests for PV pools, EV charger pools, and battery pools to reflect the new default power reset behavior; passing a new default_power parameter through several power management classes; and modifying both the ShiftingMatryoshka and Matryoshka algorithms to return default power values when no valid target power can be calculated.
Reviewed Changes
Copilot reviewed 11 out of 11 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| tests/timeseries/_pv_pool/test_pv_pool_control_methods.py | Refactored report assertion and receiver handling for default power resets in PV pools. |
| tests/timeseries/_ev_charger_pool/test_ev_charger_pool_control_methods.py | Updated expected power values for EV chargers to reflect default power (max consumption). |
| tests/timeseries/_battery_pool/test_battery_pool_control_methods.py | Modified battery pool tests to verify that default power (zero) is applied on proposal reset. |
| tests/actor/_power_managing/test_shifting_matryoshka.py and test_matryoshka.py | Adjusted expected outcomes in the Matryoshka algorithm tests to match default power behavior. |
| src/frequenz/sdk/microgrid/_power_wrapper.py, _shifting_matryoshka.py, _matryoshka.py, _power_managing_actor.py, _base_classes.py, _data_pipeline.py | Introduced and propagated the default_power parameter and logic to reset target power when proposals are withdrawn. |
Comments suppressed due to low confidence (1)
src/frequenz/sdk/microgrid/_data_pipeline.py:125
- Double-check that setting DefaultPower.MIN for PV inverters complies with the PR description (i.e. PV default power should be the minimum production power). Ensuring consistency with both the documented behavior and external documentation will help avoid confusion.
default_power=DefaultPower.MIN,
Signed-off-by: Sahas Subramanian <[email protected]>
| async def _recv_reports_until( | ||
| self, | ||
| bounds_rx: Receiver[EVChargerPoolReport], | ||
| check: typing.Callable[[EVChargerPoolReport], bool], |
There was a problem hiding this comment.
nitpick: check -> predicate? (oh, I see this is in tests, so nitpick³)
github-project-automation
bot
moved this from To do
to Review approved
in Python SDK Roadmap
|
One more thing, maybe this is worth mentioning in the docs too, in the part explaining the power manager and how proposing power works. |
Signed-off-by: Sahas Subramanian <[email protected]>
True, I've added a new commit. |
github-project-automation
bot
moved this from Review approved
to Done
in Python SDK Roadmap
2 participants
Previously, when all proposals for a set of components were withdrawn with
pool.propose_power(None), the power manager would just stop setting powers for those components, and the last set powers remain until they get reset by the API service or some other process. This is not desirable, and instead the components should get set back to their respective default powers, according to the table below. This is implemented in this PR.