Soft-sensing of NOx content in power Station based on BP Neural Network, RBF Neural Network and PCA-RBF Neural Network

Abstract

The timely and accurate measurement of the NOx content of the power plant denitrification system is very important for the precise regulation of the amount of ammonia and the control of the NOx emission. In this paper, we use 660MW coal-fired units as the research object. Combined with consideration of the poor accuracy and hysteresis of NOx measurement, the prediction models are established respectively by using RBF neural network, BP neural network and PCA-RBF neural network. The PCA-RBF model is based on principal component analysis (PCA) and RBF neural network. In order to improve the accuracy of the NOx soft-sensing model, this paper simplifies the sample model by optimizing the data samples, reducing the redundant information, and using dynamic neural network to achieve more accurate and faster training model. The results show that RBF neural network is superior to BP neural network in prediction accuracy and speed, while the improved and optimized PCA-RBF model has better prediction accuracy than the RBF neural network model. At the same time, this paper provides a theoretical basis for the real-time and accurate measurement of NOx content in power station, and provides some reference for improving the measurement and control of NOx in the actual production process.