After the processing capacity of an ethylene plant in Sinopec was expanded to 1.1 million tons per year, the vacuum degree of its cracking gas compressor unit decreased. The real-time operation parameters related to the condenser in different time periods were exported from the real-time database, python programming language was used to classify, screen and clean the big data of chemical production, the correlation was analyzed through mathematical statistics, and a three-dimensional scatter diagram was made. A multiple linear regression model of exhaust pressure was constructed based on historical data, and its accuracy was verified.