Nonlinear steam generator feedwater system identification based on wavelet neural network

Steam generator is the key equipment in the large and complex nuclear power plant, and its water supply system has the characteristics of time-varying, large delay, nonlinear. In this paper, a segmented linearized model based on wavelet neural network is developed for a steam generator feedwater system of a nuclear power plant. All the data of the automation stages of the steam generator, which are steam flow rate, feedwater flow rate and liquid level from 20% load to 100% load, are collected and divided into six stages based on the characteristics of the data. For the first four stages of data, the feedwater system was modelled using the Hammerstein-Wiener method based on wavelet neural network, which was used as the nonlinear input and nonlinear output module of Hammerstein-Wiener. And for the remaining two stages of data, the nonlinear autoregressive models based on with exogenous (ARX) inputs model approach was modelled using nonlinear ARX as the underlying framework of wavelet neural network. The proposed model can be used for model predictive control to switch between different models based on load changes. Through the verification of simulation experiments, the built model can accurately reflect the trend of actual data, and the predicted values have high accuracy and reliability.