A calibration procedure for simulation models of rural residential buildings using monthly energy bills: A case study in Zhejiang, China

Building energy simulation (BES) is crucial for planning energy-efficient retrofits in rural Chinese residences, yet its accuracy is often limited by insufficient calibration data. This study develops a practical and efficient model calibration procedure that combines monthly electricity bills with a multi-objective genetic algorithm to improve BES reliability under low-data conditions. A case study on a typical rural dwelling in Zhejiang Province was conducted. Nine key parameters—relating to envelope performance, equipment efficiency, and heating schedules—were selected to capture major modeling uncertainties. Calibration followed ASHRAE Guideline 14, targeting a Normalized Mean Bias Error (NMBE) of ≤5% and a Coefficient of Variation of the Root Mean Square Error (CV(RMSE)) of ≤15%. Results showed that NMBE was reduced from 12.775% to 0.108%, and CV(RMSE) from 21.294% to 11.054%, confirming the procedure’s effectiveness. The proposed approach enables accurate BES calibration using low-resolution data, bridging the gap between limited field data and the need for reliable energy modeling. It offers a scalable, cost-effective solution to support large-scale retrofit planning in rural settings.