From Cooling Load to Carbon Ledger: A Cloud-Native Building Operating System Evaluated Across 68 Buildings in a Hot-Arid Climate

Abstract

Commercial building portfolios in hot-arid economies face an unusual combination of pressures. Cooling can account for as much as seventy to eighty percent of a building's electricity demand, energy and carbon disclosure obligations are tightening, and the building automation estate that should provide the evidence base remains fragmented across proprietary, single-vendor controllers that do not interoperate. Conventional building management systems were engineered for supervisory control of heating, ventilation, and air conditioning, not for portfolio-scale measurement, analytics, and environmental, social, and governance (ESG) reporting. This paper addresses that gap through a design-science study. The contribution is BuildOT One, a cloud-native Building Operating System (BOS) reference architecture that organises device, edge, connectivity, time-series persistence, digital-twin, intelligence, and application concerns into independently deployable layers governed by a multi-tenant security model. The architecture is instantiated as a production platform and evaluated against six months of operational telemetry from a live deployment across a large managed real-estate portfolio in Dubai, comprising sixty-eight monitored buildings on twenty-five sites and four hundred and twenty-nine connected meters and sensors spanning electrical, thermal (cooling), water, on-site solar, and indoor-air-quality streams. Approximately seven million raw readings were ingested and reduced to verified monthly consumption using a reset-filtered register-difference method. The evaluation quantifies the platform's ingestion reliability, recovers a clear cooling-driven seasonal load signature (portfolio electrical demand rising from 3.41 GWh in February to 5.25 GWh in May), characterises on-site solar generation, and reports reading-weighted indoor-environmental-quality compliance, including 99.2 percent of carbon-dioxide readings below 1000 ppm. We further show that robust ingestion against real-world data pathologies, meter resets, register overflow, and export truncation, is a first-class operational capability rather than a preprocessing afterthought. The results demonstrate that a layered, cloud-native BOS can serve as a single measurement and intelligence substrate for energy, cooling, and ESG functions across a heterogeneous portfolio, and they surface the floor-area and baseline data that future savings-attribution studies require.