Abstract: To accurately and quickly identify the potential risk status patterns of flue gas turbines, we propose a novel potential-risk-status-pattern recognition method based on the differential evolution-extension neural network (DE-ENN). Through a theoretical analysis of the operation of the flue gas turbine, we establish a matter-element model of the flue gas turbine to determine the feature vectors and potential risk levels. Furthermore, we apply the differential evolution theory to extension neural networks and propose a solution to automatically tune the learning rate and weighted coefficients. We then present a new complete DE-ENN algorithm. In addition, we test a number of UCI standard data sets to verify the effectiveness of the proposed method. Finally, we apply the DE-ENN to the potential-risk-status-pattern recognition of the flue gas turbine. The experimental results show that the proposed DE-ENN has the advantages of requiring less learning time, higher accuracy, and less memory consumption. Moreover, the results also show that the method has better generalization ability.