Prediction of Wastewater Quality in a Certain Aquaculture Farm Based on Autoformer

The research and development teams from the School of Computer Science and the School of Resources and Environment of our university have used the BIE (Bio oxidative Information Environment) aerobic microbial enhancement system on a sewage treatment project in a certain area of Hubei province, the use of sensors and automated control technology for online cultivation of microorganisms. However, during the processing of this system, there are problems such as severe manual dependence and delayed acquisition of water quality parameters. In response to the above issues, this article predicts water quality parameters based on an improved model of the Transformer model, the Autoformer model. Firstly, the Autoformer model utilizes autocorrelation mechanisms to achieve efficient sequence level data connections and expand information utility. Then, based on the different time points in different cycles, a Time-WeightMAE loss function is designed to reduce the impact of outliers on model training. MAE and MSE are used to evaluate and explore the model by comparing with the Long Short Term Memory Network (LSTM) model, CNN-LSTM model, and Informer model, The MAE and MSE values of the Autoformer model have decreased by 10.8% and 22% respectively compared to the Informer model, which can be effectively applied to microbial culture systems and improve parameter optimization capabilities