ML-only empirical model

ML-Only Boiler Power Prediction Demo

Learning heating-cycle patterns from historical power data

Electrical input0 kW

Cold inlet18 °C

Previous draw-offDraw-off 70%

ML predicted0 W

HeaterOFF

Cycle time0.0 min

Draw-off eventComplete

Physical model vs ML-only power prediction

The selected data-driven profile is compared directly with the physical reference.

Baseline: binned average profilePhysical referenceCurrent time

The ML model remains data-driven; the physical reference starts from the usage-derived estimated temperature.

Previous hot water draw-off [%]70%

ML methodBaseline

Speed0.0 min

CyclePaused

Used to estimate the reheating cycle after a prior draw-off event, not continuous outlet flow.High draw-off · 60-80% historical bin

Outside dense training range

Prediction may look precise but may not be meaningful.

12historical training cycles

Current difference0 WMean absolute difference: 0 W

Historical heating cycles learned by the ML model

Historical cycles explain what the ML model learned from measured data.

Show rare / excluded profilesSelected: 60-80%

Baseline model

Selected ML method performance

Test MAE [W]-

ON/OFF accuracy [%]-

Both models are data-driven. They can only learn from situations represented in the training data.

Method	MAE [W]	ON/OFF accuracy
Baseline	-	-
Random Forest	-	-

Metrics are calculated on the newest 20% of representative cycles, split by cycle start time.

Physics-informed data filter

Elnett EUN 5 reference: 5 L, 2 kW, measured 18 °C inlet and 59 °C maximum hot-water temperature.

Representative-only ML training

Total detected cycles-

Representative used for ML-

Rare / possible draw-off-

Implausible / excluded-

Ideal full heat-up energy-

Ideal full heat-up time-

Advanced physical reference parameters Elnett EUN 5 defaults

Water volume [L]

Heater power [W]

Cold inlet temperature [°C]

Estimated physical start temperature [°C]

30.3 °C

Hot water / target temperature [°C]

Hysteresis [°C]

Ambient temperature [°C]

Heater efficiency [%]

Heat loss coefficient [W/K]

The reference is integrated with a one-second step using energy balance, heat loss, and thermostat hysteresis. Continuous outlet flow during reheating is zero by default; prior hot-water draw-off is represented by the calculated lower start temperature.

ML demo self-check: pending

Physical model vs ML-only power prediction

Outside dense training range

Historical heating cycles learned by the ML model

Selected ML method performance

Physics-informed data filter

Want to compare with the physical model?