The converter valve cooling system is an important part to ensure the stable operation of the converter station. The spray system continuously wets the surface of the coil by spraying the circulating water to achieve the effect of cooling and heat exchange. Circulating water is prone to breeding bacteria and al-gae and scaling, which seriously affecting the heat exchange efficiency of the converter valve cooling system. In this regard, a method for evaluating the automatic dosing state of the ultra-high pressure converter valve cooling system based on XGBoost is proposed. First, for the main components of the converter valve cooling system, 10 characteristic indicators such as inlet valve pressure, inlet valve temperature, spray water conductivity, and valve hall temperature are extracted. Then the preprocessed data is imported into the XGBoost classifier for training, realizing the parameter optimization through grid search. Finally, the field data of a converter station in Jiangsu Province is utilized to verify the ef-fect of the model. The results show that the model's evaluation accuracy rate reaches 92%.