DESS and consumption forecast

Been running a Victron Multiplus-II 3000 in my tiny house setup for about eight months now and recently started experimenting with DESS for smarter charge/discharge scheduling. Overall it's promising but I've got a nagging question about how it handles consumption forecasting.

My situation: I've got a Fogstar Drift 100Ah LiFePO4 bank paired with a small Renogy PV array (about 800W), and I'm on a completely off-grid site in a fairly cloudy part of the UK — so solar yield is genuinely unpredictable day to day.

DESS seems to pull in weather/irradiance data fine, but I'm unsure how it's generating the consumption side of the forecast. Is it just averaging historical load data from Venus OS, or is there something more sophisticated going on? The reason I ask is that my usage patterns are quite irregular — some days I'm working from home with laptops and monitors running all day, other days I'm barely there. It feels like DESS might be optimising based on a consumption curve that doesn't really reflect reality for a low-load, variable-occupancy setup like mine.

Specifically:

• Can you adjust or override the consumption forecast manually?
• Does it eventually "learn" irregular patterns, or does it just use a rolling average?
• Anyone found it makes poor decisions (e.g. discharging too aggressively at night) when actual consumption diverges significantly from the forecast?

Would appreciate any input from people who've been running DESS longer than me, especially on smaller off-grid systems rather than the larger grid-tied setups it seems more commonly discussed for.

@Thommo eight months in with a Multiplus-II and already digging into DESS — that's a solid progression, well done for pushing into that territory rather than just leaving it on defaults.

DESS really does reward the effort once you get your consumption baseline dialled in. A few things worth noting from my narrowboat setup:

• The forecast accuracy improves noticeably after a few weeks of logged data
• Make sure your grid codes are set correctly for the UK (G98/G99 depending on your export situation) — it affects how DESS models available capacity
• Pairing it with a decent Cerbo GX dashboard view helps you spot anomalies quickly

What battery chemistry are you running? Fogstar Drift cells behave quite differently to factory BMS units in how they report SOC, which can throw the scheduling off initially.

@Thommo what consumption forecast method are you using in DESS? That's where I keep hitting walls with my garden office setup — the auto-forecast seems to completely ignore the fact that I run power tools sporadically, so it keeps undersizing the scheduled charge windows.

Curious whether you've found manually inputting a consumption profile actually makes a meaningful difference to how the system schedules discharge, or whether it just smooths things out cosmetically in the graphs.

Also — are you on a time-of-use tariff feeding into the DESS calculations? I'm on Octopus Go and trying to figure out if DESS is genuinely optimising around the cheap rate or just paying lip service to it. The Victron documentation is characteristically vague on the actual algorithm.

@DodgyMechanic same here initially — the consumption forecast was doing my head in for ages.

What worked for me was letting DESS run in observe mode for a solid 2-3 weeks before trusting its forecasts at all. The learning period matters loads, especially if your usage pattern isn't super consistent.

My tiny house has pretty predictable loads apart from when I'm charging the EV overnight — that used to throw DESS completely off. Ended up scheduling EV charging as a fixed load manually in VRM so DESS could plan around it properly.

@Thommo curious what battery chemistry you're running? Found the discharge limits in DESS behave quite differently depending on whether you've got LFP or something else. Made a noticeable difference when I dialled mine in properly with the Fogstar cells.

@DodgyMechanic and @ExChippie30 — curious what you settled on in the end.

I've been running DESS with the "scheduled charges" approach rather than leaning heavily on the consumption forecast, partly because my usage in the Lakeland cottage is so weather-dependent. Electric blankets, dehumidifier, the lot — it swings massively between seasons.

What I found helpful was feeding DESS a slightly pessimistic consumption figure rather than the average. Means it occasionally leaves a bit of capacity on the table, but you avoid those nasty surprises where it's discharged too aggressively ahead of a cloudy run of days.

Also worth checking your VRM portal history — if you've got eight months of data like @Thommo, you can pull seasonal patterns and manually adjust your forecast profiles accordingly rather than relying purely on the automation. Takes a bit of fiddling but pays off.

@DodgyMechanic the consumption forecast bit genuinely feels like Victron designed it specifically to humble overconfident engineers 😄

On my narrowboat I use the historical average method rather than anything fancy — basically DESS looks at the past 7 days and extrapolates. Works surprisingly well once you've got a decent baseline logged.

Key thing I