Neuro-Fuzzy Modeling of Heat Recovery Steam Generator

Abstract

In this paper, an application of dynamic neuro-fuzzy systems is presented for modeling the subsystems of the heat recovery steam generator (HRSG). The dynamic neuro-fuzzy models were developed based on the formal NARX models topology. The clustering techniques were employed to define the structure of the fuzzy models by dividing the entire operating regions into smaller subspaces. The optimal cluster centers and corresponding membership functions are captured by FCM, where the parameters of consequent were adjusted by recursive LSE method. A comparison between the responses of the proposed models and the responses of the plants ware preformed, which validates the accuracy and performance of the modeling approach. represented in a network structure. The learning techniques for neural network can be applied in order to tune the parameters of the fuzzy models (8). In ANIFS structure presented by Jang (1999), the number of fuzzy rules is equal to the product of number of membership functions and the number of inputs (9). In some cases, the required number of fuzzy rules to cover entire input spaces is very large, which causes the training process becomes time consuming or practically impossible. In order to reduce the number of fuzzy rules without accuracy losses, the fuzzy c-means (FCM) clustering approach was proposed to define the structure of fuzzy systems (10). In this paper, a combination of fuzzy c-means clustering and least square techniques are employed to identify the parameters of membership functions and fuzzy rules in a multi-input single-output (MISO) TSK type fuzzy inference systems (FIS). The FCM clustering is first employed to extract the number of fuzzy rules and membership functions for the antecedents. Then, the parameters of consequents are defined for model based on a given set of input/output data. The accuracy of developed models is validated by performing a comparison between the responses of developed models and the experimental data. In next section, a brief description of the plant that consists of a general view of the power plant and its subsystems is presented. Inputs and outputs to the subsystems are also specified in this section. The neuro-fuzzy model based on the experimental data and structure of recurrent model and simulation result is presented in Section IV. In addition, a comparison between the responses of the proposed models with the responses of the real plant is presented to validate the accuracy of the developed models.