NOx prediction of gas turbine based on Dual Attention and LSTM

Aiming at the difficulty in predicting NOX emissions products of gas turbines in power plants, this paper analyzes the long-term operating parameter data during the combustion process, and proposes a dual attention long short-term memory (DA-LSTM) gas turbine NOX prediction model. Decompose time series by singular spectrum analysis (SSA) to remove noise in data acquisition, apply Light Gradient Boosting Machine (LightGBM) method to select features, and the dual attention mechanism is used to establish a time window to assign weights to input features and times. The model in this article is tested on the UCI public data set, and compare with related models. The simulation result shows the model can achieve effective prediction on NOX gas emissions during the combustion process.