An Improved Fuzzy RBF Based on Cluster and Its Application in HVAC System*

To solve the existing problems in modeling the HVAC (heating ventilating and air-conditioning) systems the paper proposed an improved algorithm based on fuzzy RBFNN (RBF neural networks). First, the sampling data of the HVAC system are processed using the improved fuzzy clustering method so as to redescribe the fuzzy space, thus obtaining the number of hidden layers and its parameters. Then the least square method is used to find the weights between the hidden layer and the output layer. At last, the parameters are revised by the steepest gradient descent method. Experiments demonstrated the proposed algorithm could accurately identify the model of HVAC systems even in the face of strong disturbance and system coupling. Moreover, the model outputs could track the actual outputs