Open AccessLoad Prediction and Control of Capillary Ceiling Radiation Cooling Panel Air Conditioning System Based on BP Neural Network
Author(s) -
Gang Pi,
Ling Xu,
Qiming Ye,
Pingfang Hu
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.179
H-Index - 26
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/769/4/042006
Subject(s) - air conditioning , artificial neural network , thermostat , control theory (sociology) , trnsys , ceiling (cloud) , environmental science , control system , radiant cooling , cooling load , automotive engineering , simulation , computer science , engineering , meteorology , thermal , control (management) , structural engineering , mechanical engineering , artificial intelligence , electrical engineering , physics
Compared with traditional air conditioning system, capillary ceiling radiation cooling panel (CCRCP) air conditioning system has the characteristics of low energy consumption, low noise and can provide comfortable indoor thermal environment. However, there are two problems: (1) the traditional PID control has slow feedback and lag due to the non-linearity, hysteresis and many uncertain interference factors of radiation panel air conditioning system; (2) At the start-up time, the sharp drop of temperature of the capillary ceiling radiant panel surface always cause the condensation problem. The solution adopted in this paper is showed as follows: At first, CCRCP air conditioning system model is established in TRNSYS, and the BP neural network is applied to the real-time load prediction control of the model. Secondly, fresh air pre-dehumidification system is adopted in the model to ensure that the panel condensation will not occur in the start-up stage. The results show that: 1. the correlation coefficient R of BP neural network for load prediction of working day is as high as 0.97, and the MSE is only 0.00425; 2. The temperature difference between the indoor temperature that is adjusted by BP neural network load prediction and the indoor design temperature during working hours doesn’t exceed 1°C; 3. Under the most unfavorable conditions with 1 h pre-dehumidification time, when the fresh air volume of pre-dehumidification is 100m 3 /h, the panel condensation will be prevented in the start-up stage.