The abnormality of valve cooling system in converter station is sometimes concealed in the fluctuation of normal monitoring signal and difficult to be identified. The adequacy of cooling capacity under some extreme conditions is difficult to assess. There are some problems such as few data sources and low information dimension, from the point of view of big data application. The corresponding algorithm model has simple logic and insufficient application of artificial intelligence technology. To solve the above problems, a cooling capacity modeling and early warning method for converter valves based on multi-source data is proposed. Based on the analysis of the operation principle of valve cooling system and the analysis of various state detection methods, the definition of cooling capacity is quantified and a multi-dimensional evaluation model is established. Based on multi-dimensional time series trend analysis and correlation analysis algorithm, a defect early warning model of valve cooling system is established. The validity of the model is verified by applying the model to the analysis of the monitoring data of valve cooling system at Suidong Station.