Research on seawater dissolved oxygen prediction model based on improved generative adversarial networks

The prediction of dissolved oxygen (DO) concentration in seawater is a mixed multivariate time series measurement task that is affected by many factors. In order to timely understand the status of seawater quality and reduce the losses caused by seawater pollution, it is of great significance to accurately predict the dissolved oxygen concentration in the water body. In this paper, a seawater dissolved oxygen prediction model MEMD-WGAN_IGP based on hybrid multivariate empirical mode decomposition (MEMD) and generative adversarial network (GAN) is proposed.The multivariate data after removing outliers are decomposed using multivariate modal decomposition, and the data are reconstructed into high-frequency components, low-frequency components, and trend terms by sample entropy, and then added to the improved generative adversarial network to obtain the final prediction results. The feasibility of the improved model is demonstrated by ablation experiments and compared with the classical time series data prediction model, the error MSE of the prediction results reaches 0.074, and the fitting degree R2 reaches 0.970, which is the best performance in the experiments, which proves that the model shows better prediction accuracy and stability in the marine data prediction problem.