Consumption forecast DESS is one hour to early

[Barbara (Barbara Jansink (Victron Energy))]

Hey @Ardano , we started training a forecasting model on heatpumps. Unfortunately, not many energy meters in VRM are labeled a heatpump, so because of this lack of data, the heatpump forecast does not outperform the generic consumption forecast.

Hopefully that will change when there are more devices the model can learn from.

[mac]

@Barbara

Unfortunately, not many energy meters in VRM are labeled a heatpump

I can understand why, i installed a victron VM-3P75CT to do some testing and using it for a heat pump did not appear to do anything. Doesn’t show in VRM dashboard, doesn’t show in widgets, doesn’t show in remote console overview.

Knowing it will be used to train a forecast model is great news and i will leave the meter connected to the heat pump for now!

If you want more meters set up for heat pumps i suggest either adding more functionality for them (easier said than done) or make people aware the meter is being used to train a model, for example in the VRM release notes.

[Ardano (Arjan)]

I would love to supply with my data but adding another VM-3P75CT for ev and heatpump is a bit too expensive for me. Especially since i already have acces to the consumption data via modbus and MQTT, is there an easy way to link it with vrm? I found this nodeRED integration: GitHub - Marv2190/Victron_DummyMeter_NodeRed: With this DummyMeter you can use every Datasource NodeRed support to publish Data to Victron Dbus (Gridmeter) . I think it would encourage a lot of people to categorize consumption if there would be an easy way to add them without a physical meter.

[bartm]

I’m willing to add my heatpump measurements if that would help. Which dbus servicename should I use for this? com.victronenergy.heatpump?

As said above, it would be helpful if the meter would be shown in the dashboard UI.

[jkoljo]

I believe my DESS might be scheduling 30 minutes early. That could explain the weirdness I am seeing with tonights charging schedules.

A series of incorrect DESS behaviours

Tomorrow is a very interesting day price-wise. Also a few interesting DESS behaviours, as you can see. For example: Why charge today during 21-22, when there is easily enough time to charge during the night with lower tariff after 22? Why do nothing tomorrow between 16-18, when obviously that time could be spent charging and selling. If you think there is not enough round trip efficiency to make it effective, the battery cycle price is very low, so a trading opportunity exists for sure! 17:45…

Another forum user had an identical looking timing issue

Charging battery during night time (lower price) and DESS selling with losses

This evening I noticed that the DESS started to charge up the battery at 21:30 even though my prices get cheaper at 22:00. The settings for this should be correct. [image] There’s a 30min offset in both evening and morning: [image] [image] @dognose , could you check this? Thank you! Kind regards, ej

[Barbara (Barbara Jansink (Victron Energy))]

Hey! We are adding widgets to the advanced page for the heatpump role as we speak. Expecting this to be there in 1 or 2 weeks.

[A-P]

Hi @Barbara - that’s good to hear! On the EV charger forecast, I added a VM3 meter a couple weeks back to measure the charger. The forecast is still quite weird for two reasons:

1. We only charge 3-5 times a month (on random days).
2. Our EV charger is ‘managed’ by our utility, so each charge sessions is broken up into multiple segments throughout the day / night (during favorable for grid conditions). These sessions look pretty random throughout the daytime, but a bulk happens overnight.

The combination of these two means the charges are quite random / unpredictable.

I was hoping that if that VRM sees a charge at 2am on Monday, and no charge the following Monday at 2am, then an EV charge would not be forecasted for the next coming Monday. Though we are still seeing ‘spikes’ in the consumption forecast.

Should we continue to wait a few more weeks, or will this behavior persist? If so, would it be possible to add an option to exclude EV chargers from the forecast?

[Ardano (Arjan)]

Hey @Barbara ,

I’m looking forward to the widgets.

I succesfully added the meters for heatpump and ev charger via MQTT and NodeRED.

They don’t show up in VRM yet (only in devices page).

Does it need to be in the dashboard to be used as data for the forcasting models?

[LiVi (Wikto)]

Dear @Barbara,

it looks like the issue is actual again, after return to Standard time on 26 Oct.

So, the consumption forecast is again an hour earlier.

@koningkronenberg @A-P may you please check your sites to get full picture?

[A-P]

Seems OK on my end, but our daylight savings time ends this coming weekend.

[koningkronenberg]

I checked this also, you are right that the forecast is shifted one hour but it is still in sync with the real consumption. So this issue is different from the initial issue. Everthing is one hour shifted. I think it has to do with the wintertime

[LiVi (Wikto)]

So, after one more week I am sure that it is the same issue as it was before.
Consumption forecast is again an hour earlier.

Dear @Barbara,
I will be very appreciated if you check my installation, because I am sure, I am not only person who has the same issue.
Please, take a glance at the picture - marked peaks, by fact, have to be an hour later.

[A-P]

I think you may be right, but I’m also thinking this will correct itself after a few weeks. Our DST ended this past weekend, and looking back at the historical consumption before DST ended, I see the figures shifted by 1 hour on the charts. Notice the x-axis shows 20:00, but the hover-over shows 21:00