NOx Emission Prediction of Thermal Power Plant Based on Improved Generative Adversarial Network pix2pix

Xinli Li

Abstract: NOx produced by combustion in thermal power plants will pollute the atmospheric environment. Designing an accurate and efficient NOx real-time soft sensing scheme is very important for energy conservation and emission reduction. A NOx emission prediction model based on pix2pix is proposed in this paper. The mutual information is used to rank the main variables affecting NOx generation in order of importance, and the relevant characteristic variables are selected to build the generative adversarial network model based on pix2pix. The parameters are optimized through the game training of generator and discriminator. Simulation verification is carried out based on the operating parameters of a 660MW coal-fired unit, and compared with the typical NOx prediction model based on long-short term memory neural network(LSTM), back propagation neural network(BPNN) and least squares support vector machine (LS-SVM). The results show that the proposed method effectively solves the prediction difficulties caused by the increase of data dimension, and has high prediction accuracy and small error. © 2023 Chinese Society for Electrical Engineering. All rights reserved.