Replication Data and Code for: Aerothermal renewable yield and seasonal performance of air-source heat pumps in subarctic climates: a Carnot-anchored field characterization across 26 units down to −42 °C

Description

This dataset provides the replication data and analysis code for the manuscript "Aerothermal renewable yield and seasonal performance of air-source heat pumps in subarctic climates: a Carnot-anchored field characterization across 26 units down to −42 °C". The package contains: (1) brand-level CAPM regression parameters for 26 commercial enhanced vapor injection (EVI) heat pump units monitored at ambient temperatures down to −42°C in Mohe, China; (2) temperature-binned COP data for all brands; (3) loss factor decomposition (cycling and defrost penalties); (4) six-model comparison R² results; (5) per-unit seasonal performance factor (SPF), aerothermal renewable share, and EU RED renewable-qualification flags (spf_renewable_by_brand.csv); and (6) Python scripts that reproduce the key tables and figures. All brand identities are anonymized (Brand A–Z) for commercial confidentiality. The raw minute-level monitoring data are not included due to proprietary restrictions, but the provided aggregated datasets are sufficient to reproduce all reported statistical results.

Steps to reproduce

**Prerequisites:** Python 3.9+ with pandas, numpy, and scipy installed. **Steps:** 1. Download and unzip the archive. The directory structure is: ``` AE_Replication_Package/ ├── data/ │ ├── capm_parameters.csv │ ├── cop_binned_by_brand.csv │ ├── loss_decomposition.csv │ └── model_comparison_r2.csv ├── scripts/ │ ├── 01_capm_fit_demo.py │ ├── 02_model_comparison.py │ └── 03_loss_decomposition.py └── README.md ``` 2. To reproduce Table 3 (six-model R² comparison): ``` cd scripts python 02_model_comparison.py ``` This reads `data/model_comparison_r2.csv` and prints the mean R² for each of the six candidate models, confirming CAPM achieves the highest mean R² (0.817) in 20/26 brands. 3. To reproduce Table 4 (loss factor statistics): ``` python 03_loss_decomposition.py ``` This reads `data/loss_decomposition.csv` and prints cycling loss (mean 0.25%), defrost loss (mean 1.32%), and total loss (mean 1.56%) statistics. 4. To reproduce the CAPM fitting procedure (Table A1 / Table S7): ``` python 01_capm_fit_demo.py ``` This reads `data/cop_binned_by_brand.csv`, fits the CAPM model per brand using temperature-binned COP data, and prints the estimated parameters (η₀, α₁, α₂, T_bp, R²) for all 26 brands. Note: Parameters estimated from binned data approximate but do not exactly replicate the minute-level results in Table A1, which were fitted on 276,211 raw records not included here due to proprietary restrictions. The binned-data fits confirm the model structure and reproduce the qualitative findings.

Institutions

• China Academy of Building Research

  Beijing, Beijing

Categories

Funders

• State Key Laboratory of Building Safety and Built Environment

  Beijing

  Grant ID: BSBE-EET2024-01

Licence