Research on soft sensor based on extreme learning machine

The traditional ELM algorithm in the hidden layer part parameters when the data sets are determined randomly and outliers are present, using the least squares method often exaggerate the singular value influence, cause the system deviation, resulting in the results of ELM network instability. In view of the above problems, an improved algorithm for extreme learning machine is proposed. By introducing the M estimation method, the specific idea is iterative weighted least squares estimation of regression coefficient, so as to optimize the residual sum of squares objective function to determine the weights of each sample to improve the robustness of traditional extreme learning machine. Finally, the data set is constructed to validate the proposed method, and the results show that the proposed method can reduce the impact of outliers and improve the prediction accuracy.