Under the “dual carbon” goals, high carbon emitting enterprises in industries, such as petrochemicals, are undergoing a low-carbon transformation and are reducing their carbon emissions, which have grown to become an important development trend. Oil and gas losses account for a large proportion of the total energy consumption in the petrochemical industry. In this study, focusing on the inability to automatically identify the high-value points of loss because of the change in the law of oil and gas loss data, a method for automatically identifying the high-value points of oil and gas loss based on the peak over threshold (POT) model is proposed. First, accounting of oil and gas losses was conducted in 35 links of an oil producing reservoir in Shengli Oilfield. Second, according to the leptokurtic distribution characteristics of the accounting data, the oil and gas loss tail data were segmented using the POT model and fitted with its probability distribution function (PDF). The corresponding high-value points of oil and gas losses in the PDF were identified according to the 3σ principle. Finally, the results determine the threshold of high-value points to be 869.34 m³/d, and the identification accuracy of high-value points to be 0.986, accuracy greater than other traditional methods. Therefore, the proposed method is conducive to the efficient development of loss control methods.