Generation and Classification Method of Coincident Design Day for Radiant Cooling Systems

Abstract

Radiant cooling systems have advantages in energy saving, thermal comfort with low noise, and space-saving. The coincident design days (CDDs) of convective cooling systems are not suitable for radiant cooling systems design. The generation of representative CDD is helpful for radiant cooling systems design and popularity. 800,000 sample rooms with radiant cooling panel (RCP) systems were generated for each typical city as a CDD dataset. Each sample room is a random combination of fourteen room characteristic parameters. The generation process requires feature importance analysis for room characteristic parameters. Extreme gradient boosting (XGBoost) was used to classify CDDs. For each category, the CDD with the highest presence frequency was selected as representative CDD. 80,000 sample rooms were generated for each typical city. Theoretical design cooling loads were calculated by multi-year hourly weather data. Peaking cooling loads were calculated by representative CDDs. The distribution range of relative deviation is relatively uniform. The correct rate of representative CDDs was used to evaluate the applicability of representative CDDs. Within the three threshold ranges of 0.75%, 1%, and 1.25%, the total correct rates of typical cities are 85.95%, 91.92% and 97.50% for Harbin; 93.19%, 97.49% and 98.75% for Beijing; 96.83%, 98.68% and 99.99% for Changsha; 99.24%, 99.93% and 100% for Guangzhou. The CDDs of the RCP systems could be used in radiant floor cooling system. Representative CDDs are suitable for radiant cooling systems design.