Abstract: In this study, we applied residual modified GM (1,1) model and grey topological prediction method to establish water quality prediction models based on the monitoring data of dissolved oxygen (DO), chemical oxygen demand (COD), biochemical oxygen demand (BOD₅) and dissolved inorganic nitrogen (DIN) of the artificial reef area in Haizhou Bay in spring (May) and autumn (October) of 2007−2017. Then we selected the model with higher accuracy to predict the variation trend of water concentration during 2018−2022. Finally, we verified the forecast results with the meaured data of 2018. The results indicate that the grey topological model has better prediction accuracy for water quality data than the residual correction GM (1,1). The predicted results are consistent with the actual values in 2018, so the model is reliable. According to the prediction results, the DO and COD can maintain good values in 2018−2022. It is shown that the construction of artificial reef has a certain restoration effect on the water environment of the sea area, but BOD₅ and DIN have a risk of exceeding the standard; and  improvement of grey topology model should be further explored.