Daily PV energy error hides the gap a HEMS has to close.

Why a HEMS needs more than a public solar curve

A public solar forecast usually starts with weather. It estimates irradiance, cloud cover, temperature and sometimes wind, then applies a simplified PV model. That gives a homeowner a useful first read: tomorrow looks sunny, cloudy or mixed. It gives a dashboard a reasonable expected-production curve.

A home energy management system has to go further. It is not just describing tomorrow. It is deciding when to charge a battery, when to protect backup reserve, when an EV charger can use surplus solar, when a heat pump or air conditioner can shift load, and when export or import timing is worth acting on. Those calls are not made in daily totals. They happen in operating intervals.

The question changes from "what will the sky do?" to "what will this home do?" Two houses under the same weather can behave differently because of roof angle, orientation, inverter limits, shading, dirt, snow, temperature, cable losses or simply the way the plant has behaved in the last few hours.

The comparison window

The analysis covers one GridPassport V1 home in Central Poland, with telemetry from January 1 through May 30, 2026. The broad public DayAhead baseline starts on January 16 and runs through May 30, 2026. In that 124-day public-only window, daily PV energy error is 21.92% for the known-geometry public model and 22.10% for the default-geometry public model.

The strict nowcast comparison uses May 15 through May 30, 2026, because that is where the public day-ahead forecast, GridPassport same-day forecast and GridPassport nowcast can be compared on the same active PV slots. In that narrower window, the public layer measures 17.2%, GridPassport same-day measures 12.8% and GridPassport nowcast measures 5.6%.

That daily energy score is useful, but it is not the whole HEMS problem. An active slot is a time interval where there is meaningful PV production to compare. This avoids letting night-time zeros make the forecast look better than it is, and it exposes timing and shape errors that matter for battery, EV and flexible-load decisions.

What that means without the acronym

The median slot miss is easier to read than a statistical acronym. Half the active slots were better than this number and half were worse. Here the typical public day-ahead miss was 1.57 kW. GridPassport nowcast brought the typical miss down to 0.44 kW, a 72% reduction.

The fixed thresholds below are secondary checks, not the main score. They show the same pattern in plainer terms: the nowcast was within 1 kW almost two thirds of the time, compared with a little over one third for the public day-ahead forecast.

PV is half the signal. Load is the other half.

A solar forecast tells the home how much energy may arrive from the roof. An energy manager still has to ask what the home is likely to consume. The operating signal is:

That surplus signal answers the questions that actually move devices. When will solar first cover the home load? When will it stop? Should the battery accept solar now, hold reserve, discharge, or wait? Should an EV charger pause, continue or slow down? Should the home import before a cloudy window?

In the full January 1 through May 30 telemetry period, the actual home had a PV-greater-than-load window on 127 days. The average window was 6.58 hours. In the strict May 15 through May 30 window, every comparable day had such a window, averaging 8.19 hours.

The missing loop

Most public PV tools are built to be helpful before they know the home. They ask for location, PV system size, orientation and sometimes tilt or losses. Good first pass. A modern home with solar, battery, EV charging and climate loads needs a second step: correction from the real installation.

The loop is straightforward. The system compares expected production with actual inverter telemetry, learns recent plant response and adjusts short-horizon decisions. A model that sees only weather can be directionally right and still make the wrong battery decision. A model that sees the home can learn when the actual roof behaves differently from the assumed roof.

Where GridPassport fits

The goal is not another solar screen. The public Solar Forecast tool turns location, PV setup and weather into a roof signal. A connected GridPassport goes further: it can learn from the inverter, forecast load, coordinate the battery and turn the signal into Home Power Automation.

Public PV forecasts predict the sky and a simplified solar system. A connected GridPassport works from the actual home. For an EMS, that distinction matters because battery and load decisions depend on timing, not only daily kWh.

Where the claim stops

The numbers come from one V1 home, one location and one telemetry period. They are strong evidence for the mechanism, not a blanket accuracy promise. A different roof, inverter, weather pattern, data quality level or load profile can produce a different result.

That boundary is not fine print. GridPassport should win trust by showing the data, naming the limits and improving the operating loop home by home.