Parameter selection for training process of neuro-fuzzy systems for average air temperature estimation

Adaptive neuro-fuzzy inference systems are used to develop the inferential sensor model for estimating the average air temperature in space water heating systems. Fuzzy inference system structure identification and parameter selection for structure training are the key factors for system performance. This paper describes grid partition based fuzzy inference system, named ANFIS-GRID. The impact of selection of proper parameters for training process using ANFIS-GRID is presented. Results demonstrate that selection of number of MFs, step size and step size increase rate affect the performance of the model.