ANFIS modeling of the PMV thermal comfort index based on prior knowledge

The purpose of HVAC system is to make occupants comfortable by adjusting the indoor thermal environment. The predicted mean vote (PMV) index is widely used to evaluate the indoor thermal comfort. However, PMV is difficult to calculate in real time as its complicated mathematical functions. Meanwhile, the physical conception of the model and the impact on the output of model by the human conditions are often neglected by PMV modeling in previous literatures. In this paper, all the six variables of PMV are considered. The prior knowledge about the current working conditions are used to build the initial T-S fuzzy model. Then the ANFIS is used to train and adjust the parameters of the fuzzy model through the existing dataset. Simulation results show that this ANFIS method which is based on prior knowledge not only keeps the physical means of this fuzzy model but also improves the accuracy. Moreover it is superior to the model which does not consider the human variables in accuracy of model. The proposed method is effective and accurate.