As cooling and heating energy consumptions of buildings are closely related to outdoor climate variations, the reliability of building energy simulation results is significantly influenced by the accuracy of weather data being used. We intend to construct a new typical meteorological year (TMY) for Taipei. As no beam or diffuse solar irradiance data have been recorded at local weather stations, a preliminary study on the influences of currently available hourly solar diffuse fraction models (DFMs) to the building cooling loads was performed. A 2.30%-5.18% range of annual cooling load variation was observed, which drove a need for searching suitable DFMs. To this end, the observed diffuse irradiance data of an in-situ experiment was compared to the DFM modeled values to identify the suitable DFM. It was found that Kuo's model, which has its coefficient been adapted to the local weather and further uses solar altitude, the daily clearness index as predicting variables, performed best and was used herein. The representativeness against the long-term climate of the three antiquated TMYs and the new one was discussed with a simulation-based comparison from 12 existing buildings. The reliability and accuracy of the new TMY in representing the local climate conditions are much improved. (C) 2020 Elsevier Ltd. All rights reserved.