Neural Network Based Prediction of PMV Index

This paper applies artificial neural networks theory to the study of indoor thermal comfort, a model for the prediction of PMV (Predicted Mean Vote) index is proposed by the use of back-propagation neural network. As PMV index is a highly nonlinear function of human being's thermal comfort influencing factors, so the calculation of PMV index involves complex iterations, using the approach presented in this paper can facilitate this process. Testing result shows that this model can give good results. The model presented in this paper is well suited for direct implementation using neural network chips, so it is potentially attractive for real-time air-conditioning control applications.