| Capillary radiant air-conditioning have been put into practical use in China for many years,but the application has been subject to some resistance in hot-summer and cold-winter zone and high humidity areas.The problem of the condensation on the capillary radiation ceiling is the key factor which restricting the application and promotion of capillary air-conditioning system.The capillary radiant ceiling+displacement ventilation air-conditioning system can solve the problem of condensation during the operation process,but for the buildings with intermittent operation of the air conditioning system,there will still be the problem of condensation of radiant ceiling in the system startup stage in the cooling season.Usually,pre-dehumidification of ventilation are used to solve this problem.However,for the time of pre-dehumidification,a fixed value of 1h or 0.5h is generally given according to experience,which lacks accurate calculation method.In order to get the optimal pre-dehumidification time,this paper takes an office building in Shanghai as the research object.Firstly,a model of capillary radiant ceiling+displacement ventilation air-conditioning system is built in TRNSYS.Then,the BP-neural network prediction model was established in MATLAB software,and the temperature and humidity of indoor and outdoor at 7:00 am every day was taken as the influencing factor to predict the optimal pre-dehumidification time every day.By simulating the established system in the TRNSYS software,the learning sample data set is obtained.The data set is used as input for learning and training the BP-neural network prediction model.When the training result reaches the pre-set value,the trained network is saved as a sample for actual prediction activities.The simulation results show that during the whole cooling season,when the room fresh air system and capillary radiant system are opened at the same time,the radiant roof will appear condensation phenomenon in the startup phase,and the working days with condensation phenomenon account for half of the working days in the whole cooling season.When the fan of new fresh is opened 1h in advance to pre-dehumidify the room,the phenomenon of radiant ceiling condensation will no longer occur in the startup phase of the system during the whole cooling season.The duration and the degree of condensation are related to the temperature and humidity conditions of indoor and outdoor of the day.By analyzing the learning sample data simulated by TRNSYS,the relationship between the optimal pre-dehumidification time and the temperature and humidity at 7:00am is obtained as follows:The higher the indoor and outdoor dry bulb temperature and relative humidity,the longer the optimal pre-dehumidification time;In all working days requiring pre-dehumidification,the longest pre-dehumidification time appeared on July 29,was 0.98h?58.8min?;The shortest pre-dehumidification time appeared on September3,was 0.12h?7.2min?.The mean square error?MSE?in the process of BP-neural network training was1.90958×10-4,training results of correlation coefficient R is 0.99906,predicted results of the correlation coefficient R is 0.99897.Both the training results and the prediction results have a high correlation coefficient R,indicating that the BP-neural network can reflect well the internal relationship between the input variables?indoor and outdoor parameters?and optimal pre-dehumidification time,and has a strong nonlinear mapping ability and high prediction accuracy.The neural network training is successful,and it can be saved as a sample to predict the actual optimal pre-dehumidification time. |
PDF Full Text Request