|
Abstract
To solve the problem that the purity of vinyl chloride cannot be measured in real time due to the strong nonlinear rectification process of vinyl chloride,a soft-sensor modele based on bat algorithm(BA) was proposed to optimize echo state network(ESN).Firstly,through the analysis of vinyl chloride rectification process,the auxiliary variables of the model were selected,and the normalized data were used as the input variables of the model.Secondly,since the weights and thresholds in the echo state network are randomly generated,which affects its generalization ability,the bat algorithm was used to optimize the output weight of the echo state network,thereby improving the convergence speed of the ESN model.Finally,the proposed BA-ESN model was used to predict the concentration of vinyl chloride,and the prediction results were compared with those of the ESN model and the BP model.The simulation results showed that the BA-ESN model has higher prediction accuracy,better generalization ability and robustness,and can meet the requirements of real-time measurement in vinyl chloride rectification process.
|
