What a gap in solar data really means

Separate three kinds of record

A power sample describes a moment. A cumulative energy counter describes an accumulated quantity reported by a device. An energy total calculated by the monitoring software estimates or sums energy across sampled intervals. These records behave differently when communication stops.

A device’s cumulative counter might continue advancing during a network outage and later provide a valid difference, if its documented behavior supports that interpretation. A monitor integrating power cannot know the exact missing curve from the last and next sample alone. Establish which kind of source you have before repairing a spreadsheet.

Calculate coverage before comparing totals

Keep coverage and energy as separate columns. Filling the missing hour with a daily average would create an estimate, and the uncertainty may be large. If you choose an estimate for planning, label its method and keep the measured result available beside it.

Look for the last valid boundary

Record when fresh observations stopped, when they resumed and which devices were affected. Did all sources disappear at once, or only a BMS? Did the page stop updating while the host continued collecting, or did collection itself stop? Those distinctions locate different problems.

Sunstead uses valid live intervals for accumulated energy and leaves unsupported measurements absent. Do not assume an uninterrupted-looking page means uninterrupted collection: inspect freshness and equipment state. The monitor cannot establish energy use during an interval it never observed merely by remaining open in a browser.

Use a smaller question when necessary

You may still compare the common observed afternoon on two days, or inspect whether a load change occurred after data resumed. State the limited window. Avoid comparing one full day with one partial day under identical labels.

Home Assistant’s Integral documentation describes energy integration as an approximation from samples and explains that sample behavior affects accuracy. That is a useful reminder that a long straight segment drawn between points is a charting choice, not evidence of a constant physical load.

Preserve gaps in the evidence

When exporting or reporting a problem, keep missing entries distinguishable from numeric zeros. Include the software version, timestamps and affected source. If a later correction reconstructs energy from a trustworthy cumulative counter, retain that provenance rather than silently replacing the original history.

For recurring gaps, count occurrences and durations over a defined period. A useful repair target is “prevent the ten-minute drop after the host reconnects,” not “make the chart look smoother.” Once the cause is addressed, check coverage again under the same ordinary operating conditions.

A report can be accurate while incomplete. Making that incompleteness visible is part of good monitoring.