Abstract: Aimed at the problem that the biochemical oxygen demand (BOD) of effluent in the sewage treatment process is difficult to accurately measure online, a soft measurement method based on the combination of random forest (RF) importance assessment and improved seagull optimization algorithm (ISOA) optimized long short-term memory neural network (LSTM) is proposed to predict the effluent BOD. The predictors affecting the effluent BOD were evaluated by the random forest algorithm, and the predictors with high importance scores were screened out as input variables of the soft measurement model.Secondary, an improved seagull optimization algorithm is presented for optimizing the LSTM network. The chaos mapping and new search mechanisms are introduced to overcome the problems of poor diversity and easy to fall into local optima of the seagull optimization algorithm. The ISOA is employed to optimize four hyperparameters of the LSTM network: the network iteration number, the number of nodes in the hidden layer, initial learning rate, and learning rate degradation factor. Finally, the proposed soft measurement models are applied to actual wastewater treatment processes. The results show that the prediction error of the model has been reduced and the prediction accuracy has been improved, achieving the accurate prediction of effluent BOD.

Key words: soft measurement model, LSTM neural networks, feature selection, ISOA, effluent BOD