Gray System Prediction Model of GUR Based on Error Feedback Compensation

The blast furnace (BF) is the key equipment for smelting, which consumes about 70% of the energy in the iron-making process. Therefore, improving the smelting efficiency in the context of global energy saving and emission reduction is crucial. The gas utilization rate (GUR) cannot be measured directly, as it is a ratio of different variables. The GUR can directly characterize the energy utilization efficiency of the BF, so energy consumption can be reduced if the GUR prediction model is established. This article proposes the gray system prediction model (GSPM) for GUR. First, the key variables that affect the GUR are obtained through analysis of the smelting process. Then, the time lags of the above key variables are determined through correlation analysis (CA). The principal component analysis (PCA) is adopted to rule out irrelevant and negatively correlated variables, as different variables impact GUR differently. Finally, the GSPM for BF GUR is established, and the feedback compensation is implemented based on the prediction error. The experimental results between GSPM with and without error feedback compensation indicate that the proposed model can reduce the maximum prediction error by about 55%. The mean squared error indices of the proposed GSPM, feedforward, and deep learning neural network (DLNN) under 100 samples are 0.251, 1.635, and 0.6247, which means that the GSPM performs better in the accuracy indicator.