The high-density tank is a key and complex link in the sewage treatment process,and the coagulation dosing process plays an important role in the high-density tank.Considering the characteristics of nonlinearity in coagulation dosing,the presence of substantial hysteresis,and the involvement of numerous uncertain factors,the goal of achieving predictive control of dosage and cost reduction is pursued.To address this,a combined prediction method is proposed,which incorporates Principal Component Analysis(PCA),Extreme Learning Machine(ELM),and the u-tilization of Long Short-Term Memory(LSTM)neural network residuals.PCA dimensionality re-duction and LSTM residual optimization can effectively improve the prediction accuracy of ELM,and optimize the model parameters to obtain the optimal method.Verification using sewage treat-ment data demonstrates that the proposed prediction method has mean absolute error of 0.14％and root mean squared error of 0.63％.The experimental results demonstrate significantly higher pre-diction accuracy for this method compared to random forest and other machine learning prediction methods.This provides a reliable basis for the prediction and control of coagulation dosage,and it holds practical application value.
维普
万方数据