基于门控循环单元神经网络的燃煤电厂脱硝系统建模研究

Modeling Study on Denitrification System of Coal-fired Power Plant Based on GRU Neural Network

采用某燃煤电厂1 000 MW机组实际运行数据,通过机理分析SCR出口NOX浓度与脱硝效率的各主要影响因素,使用门控循环单元神经网络建立SCR出口NOX浓度和脱硝效率预测模型。预测结果表明建立的SCR出口NOX浓度和脱硝效率预测模型的精度高于传统的RBF、LSSVM、RNN和LSTM模型,分别达到99.52%和99.63%。

The historic operating data of a 1 000 MW unit coal-fired power plant is used to analyze the main factors that influence the SCR outlet NOX concentration and denitration efficiency. A GRU neural network is used to predict the SCR denitrification outlet NOX concentration and denitration efficiency. The prediction results show that the GRU-NN model has higher accuracy than the conventional models such as the RBF, LSSVM, RNN and LSTM models. The predicted SCR denitration outlet NOX concentration and denitration efficiency are as high as 99.52% and 99.63%, respectively.