Dynamic ESS: Unpredictable Charging, Unnecessary Grid Imports

Hello, Victron community!

I have the following system configuration

Inverters: 3 × Victron MultiPlus-II 48/5000, three-phase configuration

Battery storage: 12 × Pylontech US5000, approximately 57.6 kWh nominal capacity

Solar array: Approximately 11.16 kWp, with east- and south-facing panelsʼ

Solar charge controllers: 2 × Victron SmartSolar MPPT 250/70

Monitoring and control: Cerbo GX / VRM

Electricity tariffs:

- Night (23:00–07:00): 2.16 UAH/kWh

- Day (07:00–23:00): 4.32 UAH/kWh

- Battery export to the grid: Not permitted

- Surplus solar export: Permitted

My goal is to charge the batteries overnight when electricity is cheaper and then use the stored energy throughout the day until 23:00, together with solar power, which is currently quite limited.

I have Dynamic ESS (DESS) enabled, but I’ve noticed several aspects of its behavior that I find confusing and frustrating.

1. DESS imports expensive grid electricity even when the batteries have plenty of charge.

For example, my battery SOC is around 80–85%, yet DESS decides to import electricity from the grid during the more expensive daytime tariff period.

Based on my own calculations, the available battery capacity should be more than sufficient to cover my household consumption until the cheaper night tariff begins at 23:00.

I don’t understand why DESS chooses to purchase expensive electricity instead of using the energy already stored in the batteries.

2. DESS does not charge the batteries sufficiently overnight.

Another issue is that DESS does not charge the batteries to a high enough SOC during the cheaper night tariff period.

As a result, the stored energy is not sufficient to cover my household consumption until 23:00, when the cheaper tariff begins again.

The batteries run low before the end of the day, and DESS switches back to importing electricity from the grid at the more expensive daytime rate.

I don’t understand why DESS doesn’t take better advantage of the cheaper night tariff by charging the batteries to a higher SOC, especially when this could help avoid expensive grid imports during the day.

And have you got all the settings configured correctly too? For DESS, it’s relevant to know, for example, that you’ve got 57kwh set.
Also, DESS isn’t as simple as: charge when it’s cheap, or charge at night. There’s often (but unfortunately not always) a pretty good reason why DESS does certain things.

Yes, all the settings are configured correctly.

However, on some days, DESS doesn’t charge the batteries sufficiently overnight, forcing me to use expensive grid electricity during the day.

On other days, it simply switches to grid power even though there is still plenty of energy stored in the batteries. I assume the algorithm is planning to use that stored energy later in the evening.

But when the cheaper night tariff begins, I sometimes still have a substantial amount of battery energy left that could have been used during the expensive daytime period instead.

This is exactly the behavior I find difficult to understand.

I’ve already tried making a small adjustment to reduce grid consumption during the day.

I divided the daytime period into hourly intervals and made each successive interval 10 cents cheaper than the previous one.

Since then, DESS has stopped importing electricity from the grid during the day. However, I’m still testing this approach to see how consistently it works.

The issue of charging the batteries sufficiently during the cheap night tariff period remains unresolved, though.

How long is your system running already?
Do you have overprovisioned mppts/inverters, or regular sized ?
Is all your solar “visible” to the system, or are there AC-PVInverters unknown?

For a smooth DESS Operation, the consumption and solar forecast are crucial. Wrong forecasts on either will exactly cause your issues: Undesired idling, not bought enough, sometimes the opposite etc. For the consumption forecast you would need a minimum of ~ 4 weeks of data to have it somewhat accurate.

The question about overprovisioning is due to that will heavily impact solar production forecasts, because the scheduler could not reliably determine what your systems actual peak-power might be (simplified).

ALL Solar needs to be system-visible for DESS. ESS will charge the battery from solar only concluded based on a negative grid measurement, but DESS will not be aware of “that”.

Also - as strange is it sounds - the scheduler is working “so dynamic” it really has struggles with a 2-price tariff. All sort of complex rules that have their justification in a high dynamic range cause strange effects with 2 prices only.

Thank you for the detailed explanation!

My system has been running for approximately 6 months.

It consists of three MultiPlus-II 48/5000 inverters, two SmartSolar MPPT 250/70 controllers, approximately 11.16 kWp of solar panels, and 57.6 kWh of Pylontech battery storage.

All my solar production is connected through Victron MPPT controllers and is visible in VRM. I don’t have any separate AC-coupled PV inverters.

I recently tried a small experiment: instead of using a single fixed daytime electricity price, I divided the daytime period into hourly intervals and made each successive interval 10 cents cheaper than the previous one.

Since making this change, DESS has stopped importing electricity from the grid during the day, although I’m still testing this approach.

However, I still haven’t managed to solve the issue of insufficient battery charging during the cheaper night tariff period.

Then you basically have good conditions for it to work correctly.

DESS only buys what it believes to be needed until the end of the planing horizon. And that could be to less, based on overestimated solar or underestimated consumption.

It will never do a shopping-spree like “oh, cheap today, push the battery to 90%, cause 5 rainy days”.

Since you have only 2 static prices, you can force the scheduler to buy more with a tiny node-red flow:

  • Imagine the scheduler thinks charging 30% battery is enough, but you know that mostly is about 10% to less.
  • With a node-red (timebased) you could raise the minimum soc by that 10% at 23. It may result in a immediate charge if the soc is slightly bellow that value now (doesn’t matter, not the purpose), but it will also raise the target the scheduler uses end of the current planning horizon by that 10%. It will adapt and re-schedule it’s desired 30% on top of the now raised minimum soc.
  • When the expensive period begins, you lower the min soc back to your usual value. Now, the scheduler has 40% budget till the evening rather than just it’s short 30%.

It basically will result in: Whatever the scheduler calculates to be needed for a certain day - you’ll force him to charge 10% more than that.

Oh, aside from tricking the scheduler a bit: Make sure you don’t use “Optimized WITH Battery-Life”.
That will raise your minimum soc by 5% (I think) everyday - and if that happens in the expensive period, it “steals” 5% the scheduler was planinng as budget to use.

optimized with battery life drains my battery steadily all the way down to the minimum SoC, so it must be something else.

Could you perhaps show all three VRM graphs for a problem day? Maybe something will stand out.
It can’t hurt to double-check the DESS settings either. Your daily consumption might also be useful, if it’s being compared with 57 kWh.

Yes, the DESS algorithms take some time to understand…absolutely right. I’ve been using it for more or less three years now, and at the moment I’m very grateful to have it. Especially when I look at colleagues who are customers of other companies.

It’s important to understand 2–3 fundamental things.

1 . Make sure the configuration is correct.

2. DESS ONLY ever considers the period for which reliable data and forecasts are available. This is usually the period for which day-ahead market prices are available. ONLY this period is considered. ONLY THIS ONE!!!

3. The system is very dynamic in the truest sense of the word. It adjusts its assessment very dynamically, at least once an hour. It does create a 24–36-hour schedule, but keeps changing it whenever forecasts, consumption, PV generation or the weather forecast change.

4. It always makes its decisions in the here and now. What would be the right thing to do from the current perspective? Many of these decisions (whether to charge, discharge, draw power from the grid or use the battery) may seem wrong or suboptimal in retrospect, but at the time the decision was made, it was logical.

It is great idea, I’ll try for this night

I have a very similar peak / off peak rate plan. My peak period is 13:00-21:00, and the rest off-peak.

You have a fairly large system and the DESS plan does not account for the DC consumption of your 3x inverter/chargers. For an extended peak period like yours, this continuous additional load means the consumption forecast will always be too low, and potentially cause DESS to start pulling from the grid during peak hours (given it under-forecasted how much energy is needed to stay above min SOC before end of peak period).

I used @dognose’s trick for a while, but recently switched techniques: I’ve found it’s best to trick DESS to target 100% everyday at the lowest rate - to avoid any on-peak consumption (assuming your battery is sufficient, and your peak/off peak times):

Set DESS in Green Mode, put your rates back to the flat on peak / off peak, and set your battery costs to zero (for testing).

  1. At 0:00 (or some other off-peak time), adjust the consumption forecast for 22:00 (last peak hour). Put a very large ‘fake’ load at 22:00 (e.g., 100kWh). This tricks DESS into thinking the entire battery will be needed by end of your peak period, so it will start charging your battery in the early morning (prior to 07:00) and then leave enough room for excess solar forecasted during the peak period.

  2. At 07:00 (start of peak period) clear that fake load at 22:00. This will allow the DESS scheduler to self-consume through the entire peak period (assuming enough battery capacity).

You can do this directly in the VRM portal to test it out, and then automate fairly easily with Node-Red.

Here’s an example on my system. Notice with the ‘fake’ 20:00 load that (a) it now plans to charge from grid in the morning, and (b) leaves enough room for excess solar, so that the afternoon SOC reaches 100%.

NOTE: with the fake load, you may find the system ‘schedules’ to consume from grid during the peak hours, but it won’t because you’ll have cleared the fake load ahead of the peak period starting.


Optional: Under Venus beta 3.90 you can force DESS to ‘self consume’ for the entire peak period using Node-Red (and ignore the VRM schedule). This is optional, as DESS should automatically replan to self-consume for the full peak period once you clear the fake load.

Optional: At 23:00, increase your ESS min SOC to 30% (so you have some buffer overnight if grid goes out), then at 07:00 decrease to 10%. So DESS can now use the entire battery to self consume during the peak period. You can also leave your ESS min SOC at 10% all the time.

Thanks, raising the minimum SOC by 30% helped, and the battery was charged.
I notice that everyone seems to be trying to make adjustments to “trick” the DESS, but it would be better if the system itself just worked properly.

I completely agree. And it works, too. The first few years, during development, were genuinely a bit of a struggle as a user, because some decisions weren’t very clear. But it has been working very well for me for the past year.

Read the manual and keep an eye on it. Double-check your prices, charging rates and basic configurations, then let it run. What I often observe here as a mostly silent reader is that the biggest problems crop up wherever people try to influence DESS separately. But of course, everyone makes their own decision.

Yes, I agree with you. However, it would be better to add features such as a minimum desired charge level, as well as a specific tariff mode focused on battery usage. I do agree that it works, but only because of manual adjustments; I had to increase the daytime tariff setting tenfold and decrease the nighttime one by a factor of nearly five.
Also, automating an increase in the minimum SoC during the night proved helpful.

It’s great when you enter the prices and don’t have to look at them again—everything just runs on forecasts.

And considering that the system is located in Ukraine—where power outages often occur following attacks—things get more complicated. After all, the goal is to achieve both cost savings and a reserve charge for cases when the grid goes down.