Answers are grounded in this paper and cite their sources.
Most rooftop photovoltaic (PV) systems report no live generation, so operators estimate fleet output from a few monitors, yet the metering needed to choose those monitors remains unclear. BriefCensus chooses permanent monitors from a short temporary metering campaign. Its forward greedy search matches the capacity-weighted ratio of the campaign set with no fitted weights, so the estimate needs no refit when a monitor fails. Its outage filter removes campaign rows with under half of the capacity reporting, because such rows otherwise reward survival during data outages. A logged confirmation run on 16 UK regions, without the filter, passed its rule at four weeks and kept 0.68 of the one-year selection gain. With the filter, added after that run, four weeks kept 0.82 on 12 regions of unread systems and cut the error by 27% against random monitors at k = 10. An Australian hold-out agreed. In the retrospective forecast experiment, the largest gains from monitor selection occurred at 0.5 h and 1 h. The simulated campaigns support testing a one-month metering campaign in the field.
No takes yet. Share an insight, caveat, or question.
Oussi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: