Data-Driven Assessment of Air-Conditioning System Sizing Using Runtime and Indoor Humidity Performance

Proper sizing of residential Heating, Ventilation, and Air-Conditioning (HVAC) systems is fundamental to achieving both energy efficiency and Indoor Air Quality (IAQ). Oversized systems often short-cycle and fail to remove sufficient latent moisture, while undersized systems exhibit extended runtimes and struggle to maintain comfort under peak loads. This study presents a data-driven framework that analyzes the correlation between system runtime and humidity factors to infer the likelihood of improper air-conditioning (AC) sizing. Using five-minute-interval data from ecobee smart thermostats collected across the United States between July and August 2022, compressor and fan runtimes were aggregated and examined alongside absolute humidity values. The analysis focused on approximately 4,500 homes identified within International Energy Conservation Code (IECC) humid climate zones (zones with A definition) which includes 1A, 2A, 3A, 4A, 5A, 6A, and zone 7, where latent cooling demand strongly influences indoor humidity conditions relevant to both IAQ and comfort. The compressor-to-fan runtime ratio (CFR) was used as an operational indicator of latent cooling balance, while the absolute humidity removal ratio (AHRR) represented dehumidification effectiveness by measuring the reduction in indoor absolute humidity during cooling operation. Correlation analysis revealed moderate positive associations between CFR and AHRR across most humid zones, indicating that runtime behavior can serve as a measurable proxy for moisture-removal performance. Subsequently, standardized runtime-humidity data were grouped using K-means clustering, with cluster validity evaluated through Asilhouette metrics. Based on the indoor relative humidity range, clusters with low CFR and low AHRR were interpreted as potential oversizing tendencies, while clusters with high CFR and low AHRR suggested possible undersized AC behavior. These findings demonstrate that the combined analysis of runtime and humidity performance enables a practical diagnostic of improper sizing tendencies without requiring direct capacity data. The proposed framework provides a foundation for data-driven HVAC diagnostics that support the identification of IAQ risks using smart-thermostat data.