The Thermal Comfort Modeling Based On ANFIS And Controller Design

In modern society, people spend a lot of time indoors. As a result, people have put forward higher and higher requirements on the quality and comfort of indoor environment. The traditional temperature and humidity control on HVAC have been unable to meet the requirements of them, so more and more people want to make their own comfort feeling as the direct reference during the design of the HVAC systems. For this reason, the research of air conditioning scheme based on the indoor thermal comfort has important theoretical and realistic significance. However, the definition of thermal comfort is differences among people while the indoor air object is also a typical nonlinear, pure time delay, dynamic time-varying systems. The traditional control methods are often difficult to tackle these problems. These problem need to be solved during the HVAC systems which are based on the thermal comfort. Concerning this issue, this paper will design the controller based on the model of indoor thermal comfort index PMV and carry out the research work systematically to find some feasible projects.This paper firstly introduce the research purpose and meaning of the thermal comfort index PMV modeling based on ANFIS and controller design. Then in view of those two parts: the thermal comfort model and air conditioning control, the domestic and foreign research situation and corresponding research methods are reviewed respectively. On this basis, the detailed researches have been carried out on PMV modeling and the design of PMV controller.For PMV modeling part, this paper proposed an ANFIS method used in the PMV thermal comfort index modeling based on prior knowledge. Here all the six variables of PMV are considered. Two human variables will be given a novel approach for detection which has been proved feasible. And utilize the prior knowledge to divide the input space while using neural network to learn parameters after building initial T-S fuzzy model based on prior knowledge which gives the model explanatory and accuracy at the same time. Finally, validate the proposed modeling method through the simulation.On the basis of PMV prediction model that has been given, this paper proposed an improvement plan that using hierarchical structure to tackle the problem that the PMV model has too much fuzzy rules. In this scheme, sensitivity analysis is used to sort the importance of each potential input variable on PMV so that we can choose and distribute them. The prior knowledge is also fully considered during initial fuzzy division. Some association analysis methods are also utilized when extracting the fuzzy rules among each sub-system in hierarchical structure. Finally, validate the proposed method through the simulation and comparison with that single stage PMV model.For the problem that the design of HVAC controller based on PMV index, this paper proposed a GPC design method with constraints. The PMV index is chosen as the controlled plant. In the design of the controller, the established hierarchical fuzzy adaptive thermal comfort model is used. And this model is divided into two parts, the soft sensor part and prediction model part. For the prediction model part, the fusion multiple model method is used to change the complex nonlinear relationship into convex combination of multiple linear model while become a time-varying parameter model which is suitable for GPC design. On this basis, The GPC controller with constraints will be designed. Among the simulation, this paper assumes the change form of the disturbance to close to the actual condition and proves the feasibility and validity of this design in the end.