Experimental design and the GA-BP prediction of human thermal comfort index

Fanger´s PMV (Predicted Mean Vote) is an import index to evaluate human thermal comfort. Three AM-101s (thermal environment analyzer) were used to get outdoor and indoor sample data, and then built a prediction model between the six impact factors and PMV index with genetic algorithm and neural network. In this model the difficult iterative calculation was avoided. The results show that MSE converges to 10 ∧ - 5 at the 68th epoch and the overall errors are controlled in 0.006. The correlation coefficient between PMV and the main three factors: temperature, humidity and air velocity are respectively 0.839, 0.791and-0.932. The first two factors have a significant positive correlation and the third one has a significantly negative correlation with PMV index. After daily variation analysis of indoor and outdoor temperature, this paper put forward air conditioning control measures with manual interference to provide support to the occurrence of intelligent air conditioner.