基于bp神经网络的热舒适性指标计算-南京师范大学学报工程技术版.pdfVIP

9 1 () Vo.l 9 No. 1 2009 3 JOURNAL OF NAN JING NORMAL UN IVERSITY ( ENG INEERING AND TECHNOLOGY EDIT ION ) M ar, 2009 基于 BP神经网络的热舒适性指标计算 张川燕, 王子介 (, 210042) [ ] , , . BP , PMV 3. Fanger, . PMV , Fanger 01026, 01113, . [ ] , PMV, [ ] TU 832 [ ] A [] 1672-1292( 2009) 0 1-0044-05 Therm al Com fort Index Computing Based on BP NeuralNetworks Zhang Chuanyan, W ang Z ijie ( S hool ofPow er E ngineering, Nanjing N orm alUn iversity, Nanjing 210042, Ch in a) Abstract: H um an. s om fort has a om pli ated non-linear relation w ith its influen ing fa tors. S in e the artifi ial neural networks an refle t the non-linearm apping relation, the neural networks an be used to om pute the om fort index. The BP neural newt orks are smi p ly introdu ed in this paper and a 3- layers BP neural network is established to ompute the om fort PMV index. The Fanger formu la is used to al u late the expe ted outpu t values of the input samp les and the net- w ork is trained by the input sam ples and the expe ted output values. The trained neural networks that have a h ieved the requested error are used to om pute the PMV index. The average deviation between the result from the non-trained sam- p les and the resu lt from the Fanger formu la is 0. 026 and them axmi um deviation is 0. 113. A ll the omputing errors are w ith in the requested s ope of engineering app li ation. K ey words: therm al om fort, PMV index, neural networks , , [ 1] . , [ 1-4] , , . [ 5, 6