Abstract
There are large amounts of data about buildings, which are often heterogeneous both within and across the scales of the household, building, and room. In this paper, a smart data layer is proposed, based on an ontology mapping, which maps the data from household electricity meters, building energy meters (ASHRAE), and room-environment data (UCI) into a common schema, in line with the REFIT and the SAREF4ENER and SOSA ontologies. The layer has two transparent micro-decision mechanisms: explicit operational rules and weighted multi-criteria decision making (MCDM). For the principal REFIT House 2 case study, 618 days of observations represent 5,138.9 kWh of which 24.6% falls within the defined peak period. The pipeline contains 1326 deferrable-appliance events (7.044 kWh, or 0.56% of peak-period energy). In the UCI case study, 395 of 990 occupied windows exceed the operational 1,000-ppm CO2 threshold (39.9%; Wilson 95% interval 36.9–43.0%), compared with 7 temperature violations (0.7%; 0.34–1.45%). Among 45,093 valid MCDM windows, 82.7% are classified as optimal, 6.9% acceptable, and 10.3% avoid. All 12,176 rule decisions contain generated justifications, establishing structural—not cognitive— interpretability. A dimensional audit also reveals that there is an unclosed discrepancy between the reported 7.044-kWh shift and 112.6-kg CO2 proxy, which makes the carbon estimate not be considered as validated. The results indicate that semantic integration can reveal actionable signals across domains, and also highlight the importance of uncertainty, dimensional checks, and deployment validation when making sustainability claims.