Comprehensive clustering method to determine coincident design day for air-conditioning system design

Accurate design cooling load can improve the investment economics, operating energy efficiency and reliability of building air-conditioning systems. Coincident design day (CDD) is applied to accurately calculate design cooling load air-conditioning system design. To ensure that the determined coincident design day can rationally represent and accurately predict future near-extreme design days, the clustering method should be used to obtain the overall characteristics of the near-design days in historical weather records. However, there are four 24 dimensional features in each near-design day, so that the near-design days cannot be clustered directly by the general clustering method. In this study, a comprehensive clustering comprised of weighting method and Hierarchical clustering is proposed to determine the CDD from the near-design days. The weighting method is used to combine the four 24-dimensional features into one synthetized feature vector. The Hierarchical clustering is applied to obtain the main cluster of the near-design days according to the synthetized feature vectors. The center vectors of the main cluster are calculated, and the near-design day with minimum distance to the center vectors is selected as the CDD. The results show that the CDD determined by the comprehensive clustering is more rational than that by direct selection. Otherwise, the comparison between the CDD and conventional design weather data show that the CDDs are not extreme as the conventional design weather data. Design cooling loads calculated by the conventional design weather data are overestimated by 0-30% for most cases and underestimated by 0-10% for the rest cases.

Automated summary

Accurate design cooling load can improve the investment economics, operating energy efficiency and reliability of building air-conditioning systems. The comprehensive clustering method proposed in this study provides a more rational approach to determine the coincident design day (CDD) and shows that the CDDs are not extreme as the conventional design weather data.