In the process of wastewater treatment, the aeration amount data exhibits large fluctuations and lacks clear periodicity, making it challenging for existing models to accurately predict aeration levels. Therefore, an optimized model was proposed, which utilized successive variational mode decomposition to extract features from the aeration data. Additionally, an improved Jackal algorithm was employed to optimize the hyperparameters of LSTM (long short-term memory) network, thereby enhancing the model's predictive capabilities. To address the complexity of real wastewater data, the successive variational mode decomposition algorithm was applied to decompose and reconstruct the original aeration data sequence. Subsequently, the LSTM networks were used to predict each sequence sequentially, and a Cauchy backpropagation learning mixed mutation strategy was implemented to enhance the Jackal algorithm for optimizing the parameters of the LSTM network. Finally, the predicted results of each sequence were recombined to obtain the ultimate prediction. Validation of this model using actual wastewater quality data demonstrates its effectiveness in improving the accuracy of aeration volume prediction. This model has promising applications and has the potential to significantly enhance the economic efficiency of wastewater treatment plants.