Flexmeasure use old data of sensor for schedule

[gianghm1408]

my project include solar,load,grid
step 1: generate solar , load forecast from 0h00 12/3/2026 to 0h15 13/3/2026 and update data to sensor with prior :0h15 13/3/2026
step 2: trigger schedule with prior :0h15 13/3/2026 . everything is ok
step 3 :generate solar , load forecast from 11h00 12/3/2026 to 11h15 13/3/2026 and update data to sensor with prior 11h15 13/3/2026
step 4: trigger schedule with prior :11h15 13/3/2026 . I see result scheduler gen for solar similar with data of solar from step 1

[nhoening]

Hi @gianghm1408

• What are you scheduling? A battery? Or just curtailing solar?
• And what did you expect to happen in step 4
• To allow us to understand it, screenshots would be helpful, but ideally a flexmeasures-client script that shows what you are doing.

[gianghm1408]

Hi
I scheduling both battery and solar

can you see picture . it's is result of step2 and 4. I see power of solar generate by scheduler bigger than data I upload in step 3 and similar with data in step 1. So I think It use old data to generate schedule for solar and battery

[nhoening]

Maybe you can add the call to the Scheduler and the flex-context and flex-model

[gianghm1408]

Triggering single schedule ...
[{'sensor': 963, 'soc-at-start': 2930.0},
{'consumption-capacity': '0 kW', 'production-capacity': {'sensor': 962}, 'sensor': 962}]
[DEBUG] Using building flex_context from server: {'consumption-price': {'sensor': 965}, 'site-power-capacity': '20000 kVA', 'relax-soc-constraints': False, 'site-production-capacity': {'sensor': 960}, 'inflexible-device-sensors': [952], 'site-consumption-capacity': {'sensor': 954}, 'relax-site-capacity-constraints': False}
[DEBUG] trigger_kwargs: 2026-03-12 11:15:00+07:00 2026-03-13 11:15:00+07:00
[DEBUG] ts_now=2026-03-12T09:45:15.489620+00:00 | end_dt=2026-03-13T11:15:00+07:00 | prior chosen=2026-03-13T11:15:00+07:00
final flex_model: [{'sensor': 963, 'soc-at-start': 2930.0}, {'sensor': 962, 'consumption-capacity': '0 kW', 'production-capacity': {'sensor': 962}}]
Scheduling job UUID: 69b895a7-9562-491b-8408-70c47722fc3c

here is the parameter which I use when trigger schedule

[gianghm1408]

Now I use external forecaster and update data to sensor load and sensor solar for next 24h after that I trigger schedule for next 24h with interval 15min.Is that method feasible?

[nhoening]

No consumption-capacity? So you can only use solar?

[gianghm1408]

sorry I don't get your point? can you tell more detail

[gianghm1408]

My script is
Step1:
I use post_sensor_data to post data from 0h15 12/3/2026 to 0h15 13/3/2026 with prior_ts = 0h15 13/3/2026 cho 2 sensor electricity-consumption (load) and "electricity-production"(solar -ID 962)
await client.post_sensor_data(
sensor_id=int(sensor_id),
start=start_iso,
duration=duration_iso,
values=vals,
unit=unit,
prior=prior_ts,
)
Step2;
I calll : trigger_schedule with start_dt = 0h15 12/3/2026 and duration =24h,prior_iso = 0h15 13/3/2026
final_flex_models:[{'sensor': 963, 'soc-at-start': 2930.0}, {'sensor': 962, 'consumption-capacity': '0 kW', 'production-capacity': {'sensor': 962}}]
server_fc : {'consumption-price': {'sensor': 965}, 'site-power-capacity': '20000 kVA', 'relax-soc-constraints': False, 'site-production-capacity': {'sensor': 960}, 'inflexible-device-sensors': [952], 'site-consumption-capacity': {'sensor': 954}, 'relax-site-capacity-constraints': False}
job_uuid = await client.trigger_schedule(
start=start_dt,
duration=duration,
flex_model=final_flex_models,
flex_context=server_fc,
asset_id=building_asset["id"],
prior=prior_iso,
)
Step 3 :
I use post_sensor_data to post data from 11h15 12/3/2026 to 11h15 13/3/2026 with prior_ts = 11h15 13/3/2026 cho 2 sensor electricity-consumption (load) and "electricity-production"(solar -ID 962)
await client.post_sensor_data(
sensor_id=int(sensor_id),
start=start_iso,
duration=duration_iso,
values=vals,
unit=unit,
prior=prior_ts,
)
Step4;
I calll : trigger_schedule with start_dt = 11h15 12/3/2026 and duration =24h,prior_iso = 11h15 13/3/2026
final_flex_models:[{'sensor': 963, 'soc-at-start': 2930.0}, {'sensor': 962, 'consumption-capacity': '0 kW', 'production-capacity': {'sensor': 962}}]
server_fc : {'consumption-price': {'sensor': 965}, 'site-power-capacity': '20000 kVA', 'relax-soc-constraints': False, 'site-production-capacity': {'sensor': 960}, 'inflexible-device-sensors': [952], 'site-consumption-capacity': {'sensor': 954}, 'relax-site-capacity-constraints': False}
job_uuid = await client.trigger_schedule(
start=start_dt,
duration=duration,
flex_model=final_flex_models,
flex_context=server_fc,
asset_id=building_asset["id"],
prior=prior_iso,
)

[nhoening]

I just saw consumption-capacity': '0 kW' in there. But actually I don't know what sensor 962 is.

I don't have time to think the whole thing through at the moment.

Thanks for the screenshots in any case, you built a nice dashboard :) But I'm not sure I get the whole picture.

Can you summarize what your goal is in aa sentence or two?

[gianghm1408]

962 is sensor for power production of solar. I add this line {'sensor': 962, 'consumption-capacity': '0 kW', 'production-capacity': {'sensor': 962}} to flex-model to expect schedule power of solar have lower limit is 0 kw and upper limit is value in sensor 962.
I want to schedule data need to be generate base on newest data of sensor. as you see in picture of step 4 .power solar generate by scheduler bigger than forecast data. it is unreasonable.

[gianghm1408]

I just see a problem . after step 2. I use command line in docker to check belief data of sensor power solar in sever. I see it have two data from two source first is me and second is scheduler. I use command to remove data from source scheduler and run step 3 and step 4. everything is ok . power solar generate by scheduler not bigger than forecast data

[nhoening]

Happy to hear you found the problem!

It seems you are figuring out a few basic concepts on your own, which is great (we realize FlexMeasures has a few complex layers). Not everyone gets this far.

1. It could be that your experience here shows a problem for beginners which we could fix, e.g. in the docs or even code. Deleting earlier schedules is of course not what you want to do in the end. @Flix6x do you know how by heart how to avoid mixing earlier schedules into new schedules? I'm not sure why it happened here.

2. I'd like understand what you are using FlexMeasures for. Maybe you can tell us here, or feel free to join the slack channel at https://slack.lfenergy.org (see also
   https://flexmeasures.io/get-in-touch/)

[nhoening]

@Flix6x do you know how by heart how to avoid mixing earlier schedules into new schedules? I'm not sure why it happened here.

[Flix6x]

I have a hunch it's because sensor 962 is used as the production capacity and as the sensor to save the resulting schedule to. This is exactly as suggested by our docs, but maybe it is problematic. Could you please try using a separate sensor to save the PV schedule to?

If that is indeed the issue, then we either need to update our docs or use exclude_source_types=["scheduler"] in our TimedBelief.search call in get_series_from_quantity_or_sensor. Maybe just for the capacity fields, or maybe for all fields, still needs thought. But let's first verify.

[gianghm1408]

Hi @nhoening , I have another question . after I trigger_scheduler and receive job_uuid. I call get_schedule for sensor and receive this error message. can you tell me what reason can be rootcause of this error?
2026-03-14 09:44:22,220 | ERROR | flexmeasures_client.client |
status: 400
headers: <CIMultiDictProxy('Server': 'gunicorn', 'Date': 'Sat, 14 Mar 2026 02:44:22 GMT', 'Connection': 'keep-alive', 'Content-Type': 'application/json', 'Content-Length': '323', 'Vary': 'Cookie')>
payload: {'message': 'No known schedule for this time period. Scheduling job failed with Exception: The job does not state why it failed. The worker may be missing an exception handler, or its exception handler is not storing the exception as job meta data.. was used.', 'result': 'Rejected', 'status': 'UNKNOWN_SCHEDULE'}.

[nhoening]

Has the job succeeded?

I.e. is there data in the UI?

You can also check the worker logs or the tasks tab in the FlexMeasures UI.

[gianghm1408]

here is log worker . In the log job 3e9fa892-e21f-4b54-abe1-db4c592f1790 is fail. I think worker is crash when run optimizer and restart
log_worker.txt

[Flix6x]

It says "Job 3e9fa892-e21f-4b54-abe1-db4c592f1790 made schedule." - why do you believe it failed?

I think this is a bug worthy of opening an issue, because the worker seems to report a schedule has been made, even though it is just starting to create one, and then reports again that a schedule has been made after it actually made one.

[gianghm1408]

and I see mem use of flexmeasure increase continously. can you tell me why?. I think it also is problem

[nhoening]

and I see mem use of flexmeasure increase continously. can you tell me why? I think it also is problem

yes, that would be a problem (not indented), but it's for now possible to restart periodically. I don't think we have seen this in our production systems yet, so the "Why" question I cannot answer. But of course this is always possible. We can create an issue for it, but for that better description & reproducability would be good.

[nhoening]

It says "Job 3e9fa892-e21f-4b54-abe1-db4c592f1790 made schedule." - why do you believe it failed?

I think this is a bug worthy of opening an issue, because the worker seems to report a schedule has been made, even though it is just starting to create one, and then reports again that a schedule has been made after it actually made one.

@Flix6x you do seem to have a better understanding than I do what this bug ticket should describe as the problem - can you create it?

[Flix6x]

See #2049. @gianghm1408 if you want the experience of creating a FlexMeasures PR feel free to contribute a PR removing the echo statement from the first occurrence of this code block in flexmeasures/data/services/scheduling.py, and adding a simple changelog entry under the Bugfixes section of v0.32.0 in documentation/changelog.md.

If not, no problem, I'll take it up when I can.