SoC drift / ghost discharge when peak shaving still present in 3.70

Right, so can anyone else confirm this is still happening on 3.70?

I've been pulling my hair out over the past couple of weeks because my SoC on the Delta Pro keeps showing what I can only describe as "phantom drain" — battery sitting there minding its own business overnight, barely any load, and by morning it's reading noticeably lower than it should be. No vampires in the static caravan (well, hopefully not).

Here's the interesting bit though — I had a play around last night and disabled peak shaving entirely (set it to only above min SoC), and the drift just... stopped. Overnight figures were rock solid. So it does seem like peak shaving is the culprit rather than an actual discharge issue.

A few questions for the hive mind:

• Is anyone else seeing this specifically on 3.70, or did it start earlier?
• Does it happen regardless of whether you're on grid-tied or fully off-grid? (I'm mostly running off solar with a Victron MPPT feeding in)
• Has anyone actually logged the raw data to confirm it's purely a SoC reporting issue rather than something genuinely pulling current?

My suspicion is it's a firmware bug where the peak shaving algorithm is confusing the SoC calculation somehow — like it's anticipating a discharge that never actually happens and updating the displayed figure prematurely.

Would love to know if others in the UK are experiencing this, especially anyone running the EcoFlow app through a less-than-brilliant rural broadband connection — wondering if that's adding any weirdness on top.

@EcoFlow_Nerd yes, still seeing it on 3.70 here. My setup is slightly different (LiFePO4 bank rather than the Delta Pro's internal cells) but the drift pattern looks identical — SoC creeping downward overnight even with zero actual load showing on the app.

What I noticed is it seems worse when peak shaving is actively configured rather than just enabled. Disabled the peak shaving schedule entirely as a test and the ghost discharge basically stopped. Not a fix obviously, just narrows it down.

Worth logging your actual cell voltages separately if you can — I use a Victron SmartShunt alongside and the discrepancy between what EcoFlow reports vs reality was eye-opening. Suggests it's a calculation/reporting bug rather than genuine parasitic drain.

Has anyone raised this with EcoFlow support directly? Wondering if there's an official acknowledgement anywhere.

Saw identical ghost discharge on my van build last winter — turned out the peak shaving algorithm was continuously recalculating load averages even when the feature was supposedly idle, essentially confusing the SoC estimator into thinking energy had left the bank when it hadn't.

Worth checking whether your calibration cycle (full charge → controlled discharge → full charge) resets the drift temporarily — if it does, that confirms it's a software estimation issue rather than actual cell imbalance.

@LindaJones69 the LiFePO4 flat voltage curve makes this especially brutal to diagnose since a 3% SoC error looks almost identical to genuine capacity loss.

My interim fix was disabling peak shaving entirely until a clean firmware drops — painful if you're on a time-of-use tariff, but at least your readings become trustworthy again.

@BoatPaddy that recalibration loop theory rings very true actually. I had something eerily similar on my motorhome build before I switched to a proper Victron SmartShunt doing the SoC tracking independently. The moment I stopped relying on the internal estimate and gave it a dedicated shunt, the ghost readings just vanished.

My suspicion with the Delta Pro specifically is that peak shaving and SoC calculation are sharing the same voltage sampling window — so when peak shaving kicks in and distorts the load curve, the SoC algorithm essentially sees a discharge that isn't really there.

Worth logging your voltage rails separately if you can. Even a cheap Bluetooth shunt would give you a second opinion to compare against. Isolating whether it's a genuine capacity issue or purely a firmware reporting bug would at least narrow it down.

@EcoFlow_Nerd welcome to the forum — great first post, really detailed fault description, exactly what we need to get to the bottom of these things.

One thing worth adding to what @BoatPaddy raised: have you checked whether your peak shaving trigger thresholds are set too close together? On my static caravan setup I found that a narrow dead-band between the charge and discharge trigger voltages caused the algorithm to essentially hunt continuously, which absolutely hammered SoC accuracy over time.

Worth logging your actual watt figures over a 24-hour period if you can — even just a spreadsheet. Pattern becomes obvious quite quickly.

Also curious: are you on a time-of-use tariff (Octopus Agile, etc.)? That interaction with peak shaving seems to introduce additional drift in some reports I've seen elsewhere.

Still seeing this on 3.70 here as well, for what it's worth. My setup is a bit different — running Victron kit rather than the Delta Pro — but the SoC drift pattern you're describing looks almost identical to what I logged back in autumn.

One thing nobody's mentioned yet: have you checked whether the drift correlates with times when the grid-assist threshold is being crossed repeatedly in short windows? On my system I noticed the ghost discharge was far worse during morning kettle-and-toaster peaks when the threshold kept toggling on/off every few minutes.

Might be worth pulling the raw logs if EcoFlow gives you access, rather than relying on the app display. The displayed SoC can lag the actual recalibration events by a fair bit in my experience.